Home Creators Posts Import Register Favorites Logout

Downloads

Content

!!PLEASE READ THE POST BEFORE DOING THE INSTALL!!ALSO NOTE: After the missing nodes download, a white screen will appear, just press F5 to refresh the browser window and it will look like it's loading for a while, that's normal, it's just downloading a bunch of stuff and additional models in the background.

It will then give a message like a bunch of nodes are still missing but when you go inside the install missijg nodes, nothing is showing. That's ok, it's just comfyui being dumb.

Just re drag and drop the workflow json file and the message will disapearIF you still get a missing node like Pulid or Reactor, it's probably linked to an incorrect insightface install: In that case you can try downloading the INSIGHTFACE_AUTO_INSTALL.bat file and put that bat file inside the "ComfyUI_windows_portable\python_embeded" folder.Then run it, take a note if you see any errors during that particular install.Once it's correctly installed, relaunch comfyui again, RE-drag and drop the workflow json file again inside comfyui and it's should be fine.IF the missing nodes are still there with an "import error", uninstall them all, restart, then reinstall the node again, then reload the workflow.IF YOU SEE THE ERROR "CUDA_PATH is set but CUDA wasn't able to be loaded" : To solve this, first go inside the ComfyUI_windows_portable\update folder and run the update_comfyui_and_python_dependencies.bat file. Then launch comfyui, click the manager button and click on “update all”.

Once everything is installed correctly, quit comfyui.

Download the CUDA_PATH_ONNXRUNTIME_ERROR_FIXER.bat file and put it inside the ComfyUI_windows_portable\python_embeded folder, then run it. Once you see the final message : “echo onnxruntime and opencv-python installed successfully” relaunch comfyui and try againLink to the ULTIMATE ALL-IN-ONE FLUX WORKFLOW V2: https://www.patreon.com/posts/115233635Hey everyone! I've created a 1-click installer for the newest my "ULTIMATE ALL-IN-ONE FLUX WORKFLOW V2". NOW WITH PULID, Face swaps, Automatic face and hands fixer, Outpainting and even image to video generation!!!! It's really really cool!You can check out the video right here: https://youtu.be/zLGthxUe2jcThe installer of course automates the entire install process for comfyUI and the models downloads!

IF YOU ALREADY HAVE COMFYUI INSTALLED:

1. Download the V2-MODELS-ULTIMATE_FLUX_INSTALL.bat and place it in the ComfyUI_windows_portable\ComfyUI folder

2. Run the bat file

3. Wait for the install to finish

4. ???

5. Profit 😎

LOCAL INSTALL1. Download the V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat

2. Run the bat file and select either option 1, 2 or 3 depending on your GPU vram and then A or B if you want to install the corresponding Schnell Model

3. Wait for the install to finish

4. The webui will launch, load the V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json file, you will see a lot of errors, then click on the manager button and click "Update comfyui". Then click on "Install Missing Custom Nodes" and click on the first small checkbox to select all the nodes and then click install. It will then ask you to restart, so click on that restart button, wait for the webui to reboot, click refresh once it's done and there you go😎

IF YOU ARE USING RUNPOD:

1. Create an account if you haven't already: Runpod

2. Click on Pod (on the left side) then click deploy

3. Choose a GPU with at least 24gb of VRAM (a cheap 3090 or an A30 are great), click here for the comfyui template (aitrepreneur/comfyui), then edit the template and choose 150gb for both the countainer and volume disk, then deploy on demand

4. Go to my pods, wait for everything to finish and then click "connect", then "Connect to HTTP SERVICE port 8888"

5. Then go inside the comfyui folder and then drag and drop the ULTIMATE-FLUX_AUTO_INSTALL-RUNPOD.sh file on the left side of the UI then click on the "Terminal" icon on the right side on the UI

6. Copy and paste these two lines then press enter:chmod +x V2-ULTIMATE-FLUX_AUTO_INSTALL-RUNPOD.sh

./V2-ULTIMATE-FLUX_AUTO_INSTALL-RUNPOD.sh

7. Wait for everything to be installed (it will take a while)

8. Go back to the port 3000 and click on the manager to update comfyui and restart it

9. Load the V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json file. Then click on "Install Missing Custom Nodes" and click on the first small checkbox to select all the nodes and then click install. Also use the command: tail -f comfyui.log

inside the workspace/logs to get the real time logs so that you always know when the install is done, It will then ask you to restart, so click on that restart button, wait for the webui to reboot, click refresh once it's done and there you go.

As always, supporting me on Patreon allows me to keep creating helpful resources like this for the community. Thank you for your support - now go have some fun😉!

Files

Comments

red baron

PuLID‼️ good guy

Mateusz Kloc

ULTIMATE-FLUX_AUTO_INSTALL-RUNPOD.sh is missing in attachments?

red baron

where is the file V2-ULTIMATE-FLUX_AUTO_INSTALL-RUNPOD.sh

LtCzr

Do i need to reinstall it from the file if i already have v1 downloaded?

ker0

the menu doesn't have a manager button

Aitrepreneur

any error message during install? It's supposed to be installed automatically. Otherwise, you can just go to the ComfyUI_windows_portable\ComfyUI\custom_nodes folder click on the folder path then type cmd press enter and in the command prompt window that just appeared you're gonna type: git clone https://github.com/ltdrdata/ComfyUI-Manager.git then press enter now if you reopen comfyui the manager will be here

Aitrepreneur

No read the post man, use the V2-MODELS-ULTIMATE_FLUX_INSTALL.bat put that inside the comfyui folder and run it

Aitrepreneur

noticed an error in the script, will update and upload once it's corrected, installing everything correctly for this v2 in runpod can be difficult

aprisma

Great Work. I only have 1 problem. "Missing Node Types "Text"". Any idea?

Aitrepreneur

install the missing node and re load the workflow I suppose

R

John Cabrer

A couple of nodes do not import. Anyone else? ReActor and PuLID

Savage

Where is the workflow JSON? When I search for it in the comfyui folder nothing shows up.

Aitrepreneur

read the post, I explain everything: Note: "After the missing nodes download, a white screen will appear, just press F5 to refresh the browser window and it will look like it's loading for a while, that's normal, it's just downloading a bunch of stuff and additional models in the background. It will then give a message like a bunch of nodes are still missing but when you go inside the install missijg nodes, nothing is showing. That's ok, it's just comfyui being dumb. Just re drag and drop the workflow json file and the message will disapear"

Aitrepreneur

any other nodes? see if there were any errors during install. you can also try uninstalling the nodes, restart and click install to reinstall them. otherwise it might have been a problem during the facexlib/insightface installation

Aitrepreneur

As soon as it's done but it might be very difficult to install everything needed to run the v2 on runpod

Lany

Hey auto install just crashes. It shows a new release of patch is available few times and then window closes. Any way I can debug whats wrong there?

Aitrepreneur

open a new cmd window and drag and drop the installer inside then press enter. This will avoid the crashing and it will give an error code we can use to troubleshoot the issue

Lany

Thanks! It gives fatal errors when directories already exists. Initially was pulid, now stuck on insightface. Will delete them manually

_J2_

Aitrepreneur

I'm modifying and testing the runpod script right now since the workflow version 2 really requires a lot of models to dl, I 'll repost it once it's working correctly again

Aitrepreneur

Lol OK no worries, huh I'm not sure there is an important step than is specifically made for the portable version (insightface install) but for the rest it should work fine

Aitrepreneur

you need to first update comfyui via the manager and the V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat is an installer that will install everything (comfyui portable included) from scratch, just in case people haven't done it already

aprisma

There is no mode missing in comfyui manager. The problem is the node is only called TEXT

Aitrepreneur

yes, you should read the post everything is written there :) Note: "After the missing nodes download, a white screen will appear, just press F5 to refresh the browser window and it will look like it's loading for a while, that's normal, it's just downloading a bunch of stuff and additional models in the background. It will then give a message like a bunch of nodes are still missing but when you go inside the install missijg nodes, nothing is showing. That's ok, it's just comfyui being dumb. Just re drag and drop the workflow json file and the message will disapear"

Christopher Carlson

I installed ComfyUI with your one click installer and loaded the new workflow V2 and after installing missing nodes three of them give me the error message: Import Failed. The three are; ComfyUI-GGUF, x-flux-comfyui, and ComfyUI-PuLID-Flux-Enhanced. What do I do now?

Aitrepreneur

you should read the post everything is written there :) Note: "After the missing nodes download, a white screen will appear, just press F5 to refresh the browser window and it will look like it's loading for a while, that's normal, it's just downloading a bunch of stuff and additional models in the background. It will then give a message like a bunch of nodes are still missing but when you go inside the install missijg nodes, nothing is showing. That's ok, it's just comfyui being dumb. Just re drag and drop the workflow json file and the message will disapear"

Christopher Carlson

It was called V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL

PapaNovember

I can't get the DualCLIPLoader(GGUF) to see the t5-v1_1-xxl-encoder-Q8_0.gguf for some reason. It throws an error on your default workflow, and trying to load it manually as the second clip only shows safetensor clip models that I have in there. I can still make the workflow go if I load something like t5xxl_fp16.safetensors, but it's weird that it's not seeing the GGUF files even though the t5-v1_1-xxl-encoder-Q8_0.gguf is in the same directory and not showing any errors on load. Any idea why it would do that?

Duncan

If I have V1 already do I need to re-download the Schnell model during the install?

Aitrepreneur

hmm ok so from complete scratch then, yeah then that's weird. Do you have other missing nodes or only those 3? This could also because the insightface install didn't work correctly. You can try using the INSIGHTFACE_AUTO_INSTALL.bat, put that bat file inside the python embeded folder, open a cmd window and drag and drop that inside then press enter, this will avoid the cmd window from disappearing so if there are any errors we can see them. This should hopefully solve the issue. Once you relaunch comfyui, go inside the missing nodes, then uninstall the ones giving you an inport error, restart, then reinstall them all. Then re load completely the json file to get rid of all the errors once there are no more missing nodes to be downloaded (read the post for more info)

Aitrepreneur

that's weird indeed, make sure that the file was dl correctly into the models/clip folder and is of the correct size (should be almost 5gb) you can also click on the very small refresh icon at the bottom of the small floating bar, this should refresh the whole ui and make all the files/model appear correctly

Aitrepreneur

If you already have the V1 you need to use the V2-MODELS-ULTIMATE_FLUX_INSTALL.bat inside the ComfyUI_windows_portable\ComfyUI folder, run it and you'll have the possibility of choosing to dl schnell or not

Stunspot Prompting

Perhaps you might copy it here? I am not actually seeing it there, sir.

latentpunk

would you know why the videoupscaler wouldn't work? the 8fps version works, but then when it comes to the interpolation part, the second video combiner never runs and the promp ends. thanks update: nevermind, it wasn't loading the upscaler because there was none. i downloaded the 4x clearreality upscaler and it works now)

Lagartija

I'm getting this warning during install - " WARNING: The script insightface-cli.exe [+ a whole bunch of other .exe's] is installed in 'F:\AIs\ComfyUI_windows_portable\python_embeded\Scripts' which is not on PATH. - Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location." will it be an issue? how do I add to PATH?

PapaNovember

yeah it is there, and 4.9 gig, and refreshing doesn't help. I've seen a few other people report this problem elsewhere so it may just be some kind of isolated glitch.

Bryan Crump

found the .json file and had to update again but finally saw where to install missing addons, at first nothing showed, then a ton of stuff popped up, so at that point now

Mr.E

[Using a nvidia RTX 2080] I've been getting this error "SamplerCustomAdvanced expected scalar type Half but found BFloat16" for PULID. Any idea on how to resolve this, AI OVERLORD?

Bryan Crump

says these nodes are missing... workflowFlorence workflowFlux Controlnet workflowNodes 3 workflowXLabs Sampler workflowAdetailer workflowoutpaintsize1 workflowoutpaintfluxend workflowoutpaintflor workflowVIDEOUPSCALER workflowvideoencodecode workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpullflo workflowpulidtxt workflowpluidlux

latentpunk

had the same problem, search for the name on google + comfyui, then download and replace it. the file seems to be corrupt

latentpunk

would you be able to make an extend for the image to video portion? that would be so cool.

todd beebe

i am getting this same thing?

todd beebe

where is this new json file?

Bryan Crump

let me know if you find the solution, i have tried rendering but keep getting missing node errors

Tinkering

General question. I have a dual 4090 system (mostly used for testing Gen_AI models) with Auto1111 Image generation is a few seconds. I am fairly new to Comfy and just installed your workflowV2. Image gen is slow. Of course, the first one is slow but all subsequent ones take a good 15 seconds. Normal?

JohnThePizzaEater

Ughhh. Like many other people, I'm having errors with missing nodes. Updating ComfyUI and going into the manager to install missing custom nodes does nothing. I even go into the menu to look at missing nodes and none are found.

DampyDan

so i had that too when it loaded up and for me what worked was closing that little window and then dragging and dropping the workflow.json file in there and it just worked

Klaus The.

You have to click the "Clear" in the UI and then reload the .json workflow file again

Bryan Crump

that seems to have brought progress, thank you

Janne Kallio

does not seem to fix it.. at least to me.. I have refreshed, clearer, restarted, reloaded and all, but still ReActor Node for ComfyUI and ComfyUI-PuLID-Flux-Enhanced are (IMPORT FAILED) .. I keep trying and report..

Duncan

Thnx!

Bryan Crump

now, I get this error when trying to render: ApplyPulidFlux Allocation on device

Bryan Crump

I got image generation to work but when i click add lora button no window opens or anything

Bryan Crump

it's my bad, sorry it's been a minute since I have done this, forgot to load them into the lora folder, lol

Bryan Crump

still need to figure this out, does it need a higher gpu?

aprisma

Amazing. Thanx so much for your work. Even if I setup my comfyui totally new using your one click installer and after your new all in one ultimate workflow V2, some of the nodes are missing. Just for example At the Group: FP8 PULID, the second node on the right side (clip, text1,text2,...text5, Conditioning) is red. Also the last node right at the end. (lora, image, image, mask, caption,data) What node could I take instead? Or what is the name of the node? I tried NEARLY similar nodes but ..... ADD: The same with FP8 IMAGE TO PROMPT: the middle node is missing but no missing node in comfy manager. What node is that? So in total except in the first workflow always one node is missing that is not listened in missing nodes :-((

Number 6

Thanks for all your hard work. I'm struggling with Pulid and getting insight-face errors. Any thoughts? I'm running on Linux. # ComfyUI Error Report ## Error Details - **Node Type:** PulidFluxInsightFaceLoader - **Exception Type:** AssertionError - **Exception Message:** ## Stack Trace ``` File "/home/ian/ComfyUI/execution.py", line 323, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "/home/ian/ComfyUI/execution.py", line 198, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "/home/ian/ComfyUI/execution.py", line 169, in _map_node_over_list process_inputs(input_dict, i) File "/home/ian/ComfyUI/execution.py", line 158, in process_inputs results.append(getattr(obj, func)(**inputs)) File "/home/ian/ComfyUI/custom_nodes/ComfyUI-PuLID-Flux/pulidflux.py", line 196, in load_insightface model = FaceAnalysis(name="antelopev2", root=INSIGHTFACE_DIR, providers=[provider + 'ExecutionProvider',]) # alternative to buffalo_l File "/home/ian/ComfyUI/comfyui-env/lib/python3.10/site-packages/insightface/app/face_analysis.py", line 43, in __init__ assert 'detection' in self.models

Duncan

This is the winner. Worked for me, thanks.

Duncan

All working now. Fantastic effort putting this all together, congrats and thanks.

Janne Kallio

I am not sure if it helps. I did it, but had to move the value below the python path... if you are on Win copy that path, then search for "edit the system environment". click it, then on bottom right click "environment variables". there on the lower part "system variables" search for "path" and double click on it. then click "new" and paste the path there. make sure it is below your python path on the list.. I am not sure if this helped, but I noticed error when starting that Insightface was not found.. it is installed, but there is a problem with the python versions.. To make it work I had to go to ComfyUI_windows_portable\python_embeded> and give the following commend: python.exe -m pip install insightface==0.7.3

Janne Kallio

I had the problem that few nodes were not installing no matter what I did. I noticed error when starting that Insightface was not found.. it is installed, but it seems there was a problem with the python versions.. (global python vs install version or something).. To make it work I had to go to ComfyUI_windows_portable\python_embeded> and give the following commend:python.exe -m pip install insightface==0.7.3 after this I got no errors.

J Rosen

Has anyone installed this onto a new runpod as it is taking up 600GB on the Pod Volume. This a lot more than what the guide says at 150GB.

Walker4k

Txt to Image error - [] [] got prompt Using pytorch attention in VAE Using pytorch attention in VAE !!! Exception during processing !!! Error while deserializing header: MetadataIncompleteBuffer Traceback (most recent call last): File "O:\Comfy_UI_V4\ComfyUI\execution.py", line 323, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "O:\Comfy_UI_V4\ComfyUI\execution.py", line 198, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "O:\Comfy_UI_V4\ComfyUI\execution.py", line 169, in _map_node_over_list process_inputs(input_dict, i) File "O:\Comfy_UI_V4\ComfyUI\execution.py", line 158, in process_inputs results.append(getattr(obj, func)(**inputs)) File "O:\Comfy_UI_V4\ComfyUI\nodes.py", line 937, in load_clip clip = comfy.sd.load_clip(ckpt_paths=[clip_path1, clip_path2], embedding_directory=folder_paths.get_folder_paths("embeddings"), clip_type=clip_type) File "O:\Comfy_UI_V4\ComfyUI\comfy\sd.py", line 437, in load_clip clip_data.append(comfy.utils.load_torch_file(p, safe_load=True)) File "O:\Comfy_UI_V4\ComfyUI\comfy\utils.py", line 34, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) File "O:\Comfy_UI_V4\ComfyUI\venv\lib\site-packages\safetensors\torch.py", line 313, in load_file with safe_open(filename, framework="pt", device=device) as f: safetensors_rust.SafetensorError: Error while deserializing header: MetadataIncompleteBuffer

Seanehty

As a general trouble shooting technique when you are having issues loading models or are missing nodes: 1) Load ComfyUI (Presumes you have run the ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat ;-) ) 2) Clear any loaded workflows 3) Go to the Manager button, click it and ensure you have updated ComfyUI 4) Restart comfyUI 5) Load the workflow you want to use like the V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json 6) Go back to the Manager button, click it and ensure you have installed all missing nodes 7) Restart ComfyUI 8) Clear and reload the workflow you want to use. Then see if you have any errors. ;-)

Daniel

just installing and my Volume is fine by now.

Daniel

feel you :D same mistake all the time for me

Alexey R

[INFO] Downloading diffusion_pytorch_model-00001-of-00003.safetensors... % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 1180 100 1180 0 0 9778 0 --:--:-- --:--:-- --:--:-- 9833 10 4761M 10 512M 0 0 22.7M 0 0:03:29 0:00:22 0:03:07 19.2M curl: (23) Failure writing output to destination

McGenie

I had a similar issue - and this worked for me.

Seanehty

Possible space issue? Permissions? Did you retry?

ker0

say thing again but like you are talking to an idiot.. cause I too am struggling to get the faceswap to work

Steve Geddes

FYI, I have no idea what I'm doing, but I got it working. I downloaded t5-v1_1-xxl-encoder-Q8_0.gguf and replaced the existing file. I downloaded the ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors and replaced it as well. Restarted ComfyUI, got the same error. Updated ComfyUI, unloaded the workflow, loaded it again and now it works.

McGenie

Also specifically for Insight: Try using the INSIGHTFACE_AUTO_INSTALL.bat, put the bat file inside the python embeded folder, open a cmd window and drag and drop that inside then press enter. For Pulid: If the install fails delete the folder "ComfyUI-PuLID-Flux-Enhanced" from ComfyUI\custom_nodes and try installing again from the ComfyUI manager.

Michael O'Neill

I really appreciate your attempts to create this one click installer, It would be awesome to include a troubleshooting guide, one of the other users suggested clearing and reloading the workflows, and that solved my issues. Would it be possible to create an uninstaller?

Marc Bridewell

I have zero knowladge of any of this stuff and wanted to try it out. How I ran the script and nothign happend I assume I need to get this ComfyUI program? Can somone point me at a guide for instillation I googled it but I dont know if websites Iare legit. I heard that some websites for stable diffusion pass have been trageting people with malware.

LL

I'm just another user just like yourself. Yes, you need the ComfyUI program to run on your computer. There are 3 methods to accomplish this and I have successfully tried all three. 1) Install ComfyUI as a standalone product. 2) Download and run another catch-all product called Stability Matrix (Github). 3) Alternative and similar is Pinokio. The latter two (I use both!) offers numerous AI programs you might find useful as you build and expand your interests in the field whether it be chat, image, sound, video and a few other related AI projects. Start with ComfyUI and add others as you progress. The advantage of the two "catch-all" is they offer internal update channels with one press from their interface. ComfyUI Standalone will offer updates if you go back into its folder and click the "Update" link (just a minor inconvenience). YOU SHOULD BE RUNNING AN NVIDIA CARD TO COMFORTABLY ACHIEVE PROCESSING! AMD is trying to get caught up with Nvidia's CUDA system which currently has the upper hand. There are some workarounds but Nvidia still rules at this point. (I'm talking a couple of minutes vs maybe an hour or so in generation time!) So much more I can say but AITREPRENEUR has many fine videos on the process since you're on his Patreon anyway. https://github.com/LykosAI/StabilityMatrix https://pinokio.computer/ https://github.com/comfyanonymous/ComfyUI (standalone) Get a Github account, too! Python will be required. And plenty of hard drive space for models. Also, Stability Matrix's system uses a shared folder system between its programs which comes in handy since one model is generally 6+GB BUT ComfyUI, Invoke, Forge, et.al., share the models (and other components downloaded) so it's not a 6+GB download for EACH program! Lots of space saving you will need if you expand!

Benoit Huguet

This is simply amazing ! Thanks a lot for putting so much hard work into such a simple plug and play solution ! Everything works perfectly in just one click :)

Marc Bridewell

thanks I got it working I have a prety decent PC with a good video card. I followed another youtube video. Now I just need to figure how to load the .json lol

Vaivij

I get the following error when I run the Outpainting workflow: CLIPTextEncode: 'NoneType' object has no attribute 'tokenize' In the comfyui.log file, I see this: [2024-11-03 18:44:22.182] Traceback (most recent call last): File "/workspace/ComfyUI/execution.py", line 323, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "/workspace/ComfyUI/execution.py", line 198, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "/workspace/ComfyUI/execution.py", line 169, in _map_node_over_list process_inputs(input_dict, i) File "/workspace/ComfyUI/execution.py", line 158, in process_inputs results.append(getattr(obj, func)(**inputs)) File "/workspace/ComfyUI/nodes.py", line 64, in encode tokens = clip.tokenize(text) AttributeError: 'NoneType' object has no attribute 'tokenize' [2024-11-03 18:44:22.186] Prompt executed in 90.20 seconds

Michael O'Neill

I have a machine with two GPUs, is there any way to create a workflow that will take advantage of this? I can only get ComfyUI to utilize one. Thanks!

Vaivij

Is it possible to add cogVideoX with motion guidance (Tora)? How can I incorporate that portion in the workflow?

Ramon Rodgers

I have an error where Face Restore Models keeps re-downloading the same files over and over again. [ReActor] Downloading GFPGANv1.3.pth to H:\[ AI ]\flux\ComfyUI_windows_portable\ComfyUI\models\facerestore_models\GFPGANv1.3.pth:

Lagartija

sometimes the download will hang - for example: 86% of 4754M 86 4109M 0 0 1185k 0 1:08:26 0:59:09 0:09:17 0, where 0 is the dl speed. is there a way to restart the dl without having to restart the whole process?

Senjka Danhieux

On Runpod, it always goes wrong with installing the custom nodes. It never works. I am following the guide, but for some reason they fail to install ... unknown error. The ComfyUI Impact Pack just stays on : Restart Required now... no matter how many times I restart.

Petri Lummemäki

I would like to keep the model files separate from ComfyUI and I have edited the extra_model_paths.yaml file so that ComfyUi finds checkpoints, clips etc. there. But can I edit the extra_model_paths.yaml file so that ComfyUI finds the xlabs, pulid, insightface etc. files from that separate location or do I have to put them in ComfyUI's internal directory?

benoit corvol

It work for me, did you try to install them one by one ?, and update all ?

Senjka Danhieux

I had to constantly close, restart, reconnect, clear, reload the workflow and then little by little I got them all.

Janne Kallio

I am getting same safetensors_rust.SafetensorError: Error while deserializing header: MetadataIncompleteBuffer when trying to do the image to video. I got everything else working. I have a feeling some models are not fully downloaded. It would be great to be able to check this somehow. When using the model downloader, it opens and runs something and I can see some red text, but it closes too fast for me to see what is was. using the loader in admin mode it seems to download again things I already got. my connection is to bad I fear something goes wrong every time is I download everything as it takes hours. EDIT: I am reading the following: The solution is to re-download or otherwise fix your safetensors file. I am investigating and for me it seem some file sizes do not match. the CogVideoX-5b-I2V/transformer folder files have different size than in: https://huggingface.co/THUDM/CogVideoX-5b-I2V/tree/main so I am now downloading the files manually and will copy them later.. lets see if that works ... -> UPDATE: That worked out! it is now working. just one file needed to be updated

kristopher ulley

first off thank so much for you hard work! This is awesome! I'm having one issue though. I seem to be able to make a image or two but it always end up freezing. I did a fresh install. I have a 4090 and i9 gpu so i should not be running out of vram. and ideas on what might be wrong? Also Would you be interested in building me a custom workflow? I would pay you of course.

Tyler Horn

Where is the json file?

Aitrepreneur

go inside the custom_nodes/ComfyUI-Manager folder and edit the config.ini file and change the last line to security_level = weak save the file and relaunch comfyui

Aitrepreneur

any errors in the cmd window? Have you tried both the Fp8 and GGUF version of the workflow?

Aitrepreneur

huh might be but not sure tbh, I personally wouldn't play with that if you don't want to spend hours troubleshooting for no reason

Aitrepreneur

you can just use the V2-MODELS-ULTIMATE_FLUX_INSTALL.bat instead once you got comfyui installed, put that inside the ComfyUI_windows_portable\ComfyUI folder and run it, it will automatically check which files are missings and only downloads what is missing

Aitrepreneur

it's normal that it downloads that file once but again and again that is weird, is that file present in the described folder?

Aitrepreneur

I don't think comfyui can take advantage of multiple gpus as far as I know, it's not a workflow issue

Aitrepreneur

Maybe because you changed something in the model selection, usually that happens when you chose the incorrect text encoder in the the dualcliploader node. The 1st clip must be the "ViT-L-14-TEXT-detail-improved-hiTGmP-TE....." and the 2nd must be your T5 xxl encoder

Aitrepreneur

can you send me some screenshots? I never saw that error. Have you tried also re drag and dropping the workflow file inside the ui and refresh?

kristopher ulley

I have been using the fp8 version so I will try the other. I also updated phyton. It will run fine as long as i don't make any changes but if i change something it will freeze.

Christopher Moonlight Productions

I'm getting a message saying it can't finish the install because PYTHON311.DLL is missing. I went and installed it and tried running it again, and it just stops. I'll try a few things but maybe there's something I'm not thinking of?

kristopher ulley

last line in dos is loaded completely numbers and then TRUE

Aitrepreneur

That is...a weird error? What installer have you used?

Christopher Moonlight Productions

I think I may have figured it out. I somehow managed to delete some of the launch files in my ComfyUI folder. Not sure how that happened but I'm re-running it.

Michael O'Neill

Any other ideas? I have a Nvidia4060 16gb and a 3060Ti.

Wanni

Hi, Thanks for the previous help. Now I'm stuck in trying to install missing nodes. I tried fixing it with the option in the manager. but they seem to come back. They look like important nodes to get start with creating amazing pictures like yours. Can you help me with this. Thanks in advance. (IMPORT FAILED) ComfyUI Impact Pack 25 (IMPORT FAILED) WAS Node Suite 137 (IMPORT FAILED) ReActor Node for ComfyUI 1459 (IMPORT FAILED) ComfyUI-PuLID-Flux-Enhanced

Janne Kallio

I see the instructions is updated. I guess the INSIGHTFACE_AUTO_INSTALL.bat does exactly this :)

Janne Kallio

What we need now is discussion on the video tricks. so far the results for me are not mind blowing. Camera moves, but there is really not much movement in any characters

BOB

Thank you for this post, which is much more detailed about potential errors and helped me resolve them more effectively. I struggled a bit with the last "ultimate v1", especially in realizing that ComfyUI can just be "dumb" at times. But now everything has gone smoothly. Great work, thanks again !

Firmy Van Bas

Thanks for that! Just did all the install locally and I'm getting the error that Reactor nodes are missing. I did the Insightface_auto_install, installed all missing nodes and updated comfyUI but I'm still getting the same warnings at the start. Any idea? thanks!

Petri Lummemäki

Ok, I'll try that then. I've tried to keep the model files separate so that I don't accidentally delete them if I have to reinstall the program, as downloading them is quite a job and some files may have been completely removed from the web.

Aitrepreneur

I explain everything in the post at the beginning: IF you still get a missing node like Pulid or Reactor, it's probably linked to an incorrect insightface install: In that case you can try downloading the INSIGHTFACE_AUTO_INSTALL.bat file and put that bat file inside the "ComfyUI_windows_portable\python_embeded" folder. Then run it, take a note if you see any errors during that particular install. Once it's correctly installed, relaunch comfyui again, RE-drag and drop the workflow json file again inside comfyui and it's should be fine. IF the missing nodes are still there with an "import error", uninstall them all, restart, then reinstall the node again, then reload the workflow.

Aitrepreneur

Blame the video model not me, that's the limit of current open source video model, and it's still one of the best open source ones that I've tried

Aitrepreneur

sorry I don't have any apple products so I don't do mac installs also comfyui only work on M1 or M2 chips I think

Aitrepreneur

any error in the cmd window? where there any error during the Insightface_auto_install install as well?

Christopher Moonlight Productions

Well, I did a clean install but now it can't seem to get these nodes... workflowsampler and seed upscaler workflowExtra nodes upscaler workflowFlorence workflowFlux Controlnet workflowImage gen sampler workflowOther Nodes workflowParam workflowsampler img workflowOther Nodes 2 workflowNegative Prompt workflowSampler 3 workflowNodes 3 workflowXLabs Sampler workflowpositive prompt guidance workflowAdetailer workflowfff workflowoutpaintsize1 workflowoutpaintfluxend workflowoutpaintflor workflowVIDEOUPSCALER workflowvideoencodecode workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpullidpara workflowpullflo workflowpulidtxt workflowpluidlux

MonoChromancer

Thank you so much! your content, tools and work is amazing. My favorite AI content channel, informative, entertaining and useful Suggestions (for future ultimate workflows): 1. Face merge: Provide X number (up to say 6 or so) of face inputs and merge them (by default proportionally based on the number of populated inputs, or alternatively you could provide a value for how much a particular face contributes) onto the face with merged traits 2. Body detection and then prompt to change (like your hands and face workflow) 3. Skeletons/pose stuff to the workflow (there's some great work out there with being able to capture poses and motion and turn them into skeletons, and then use those poses/skeletons to generate motion/ video 4. Optional sound effect (you can drag and drop a new one but have a default) for when a single operation is complete, and a different one for when the entire queue is done. I made a node back in the day to do this but had to keep readding it with updates and stopped but it was very useful to let me know with a long queue when all my work was complete while I had it minimized. 5. optional "Save workflow to generated image" 6. Expose the seed used and an option to lock the seed, increment it per generation, or randomize it per generation 7. Optionally create a folder for better organization of files, by default it uses the date if no override is added to the field 8. the ability to have a global override for all workflows (gets complicated in the workflow cases where you use 2 like the SD then Flux workflows) for the model being used. this can be important since some models are free or free up to a point and some models are only free for non commercial use and I wouldn't want to set one workflow to one if I am in a position where I am either setting up or transitioning between work that might need to care about commercial vs not and don't want to have to manually go through each workflow and make sure it is correct for each swap. thank you so much, please keep doing all the incredible work that you are doing I have a few bugs but Ill post in a separate comment for clarity

MonoChromancer

I am getting the same error I believe here are the errors I am getting: (I tried a few steps of uninstalling conflicted stuff and uninstalling and reinstalling these nodes as well as "try fix" and I tried following the steps in the link in the error. Comfyroll Studio v1.76 : 175 Nodes Loaded ------------------------------------------ ** For changes, please see patch notes at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/blob/main/Patch_Notes.md ** For help, please see the wiki at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/wiki ------------------------------------------ [comfyui_controlnet_aux] | INFO -> Using ckpts path: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts [comfyui_controlnet_aux] | INFO -> Using symlinks: False [comfyui_controlnet_aux] | INFO -> Using ort providers: ['CUDAExecutionProvider', 'DirectMLExecutionProvider', 'OpenVINOExecutionProvider', 'ROCMExecutionProvider', 'CPUExecutionProvider', 'CoreMLExecutionProvider'] DWPose: Onnxruntime with acceleration providers detected # 😺dzNodes: LayerStyle -> Cannot import name 'guidedFilter' from 'cv2.ximgproc' A few nodes cannot works properly, while most nodes are not affected. Please REINSTALL package 'opencv-contrib-python'. For detail refer to https://github.com/chflame163/ComfyUI_LayerStyle/issues/5 [rgthree-comfy] Loaded 42 fantastic nodes. 🎉 WAS Node Suite: OpenCV Python FFMPEG support is enabled WAS Node Suite Warning: `ffmpeg_bin_path` is not set in `H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui\was_suite_config.json` config file. Will attempt to use system ffmpeg binaries if available. WAS Node Suite: Finished. Loaded 218 nodes successfully. "Your time is limited, don't waste it living someone else's life." - Steve Jobs Import times for custom nodes: 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Logic 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Universal-Styler 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-mxToolkit 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_JPS-Nodes 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-various 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_UltimateSDUpscale 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Chibi-Nodes 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\Derfuu_ComfyUI_ModdedNodes 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Custom-Scripts 0.0 seconds (IMPORT FAILED): H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node 0.0 seconds (IMPORT FAILED): H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Frame-Interpolation 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\x-flux-comfyui 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-KJNodes 0.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy 0.1 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux 0.1 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Long-CLIP 0.1 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_Comfyroll_CustomNodes 0.1 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-dream-project 0.3 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager 0.4 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2 0.4 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use 0.4 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack 0.6 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-tensorops 0.7 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-VideoHelperSuite 1.0 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper 2.4 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui 2.9 seconds: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_LayerStyle Starting server To see the GUI go to: http://127.0.0.1:8188 FETCH DATA from: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE] [ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node. Traceback (most recent call last): File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\server.py", line 564, in get_object_info out[x] = node_info(x) ^^^^^^^^^^^^ File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\server.py", line 531, in node_info info['input'] = obj_class.INPUT_TYPES() ^^^^^^^^^^^^^^^^^^^^^^^ File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in INPUT_TYPES "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "pathlib.py", line 931, in iterdir FileNotFoundError: [WinError 3] The system cannot find the path specified: 'H:\\AI\\OldFlux Sept2024\\ComfyUI_windows_portable\\ComfyUI\\models\\LLM' Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI [] [] FETCH DATA from: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE] [ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node. Traceback (most recent call last): File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\server.py", line 564, in get_object_info out[x] = node_info(x) ^^^^^^^^^^^^ File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\server.py", line 531, in node_info info['input'] = obj_class.INPUT_TYPES() ^^^^^^^^^^^^^^^^^^^^^^^ File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in INPUT_TYPES "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "pathlib.py", line 931, in iterdir FileNotFoundError: [WinError 3] The system cannot find the path specified: 'H:\\AI\\OldFlux Sept2024\\ComfyUI_windows_portable\\ComfyUI\\models\\LLM' Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: H:\AI\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI

Janne Kallio

well.. I was not blaming you :) I am sure there are some prompting tricks and setting to play around with.. or we wait for better models that are for sure on their way :)

MonoChromancer

I saw someone post a similar issue to mine (the Reactor nodes are missing issue) so I am leaving that one out from this post. Issue 1: When I use the 1 step installer (no matter which I try) I don't end up with the comfyUI manager for some reason, I thought this was included, so I end up using an old install rather than chasing down the manager install each time. this has happened with each of the ultimate installers. But I think this causes me to just smash the install over the previous which gives me the manager but then I end up with node conflicts that I have to manually sort out. issue 2: When it asks if I wan the schnell install I am always questioning and then having to go back to your videos to check is it "so you -only- want schnell instead?" or is it "do you -additionally- want schnell?" In my case I would prefer the dev one most of the time but I do optionally want to swap to schnell in some cases so it would be grat to have the wording a bit clearer (or provide several options, dev only, schnell only, dev + schnell)

Christopher Moonlight Productions

Okay, I managed to update them by loading OTHER workflows and updating them through there but I could not update them with this workflow. I have no idea why that's happening but it worked. lol. Thank you for the workflow.

Adam Booth

After following the instructions (I think) when it loads lots of the tools still have red boxes. For example Face swap 1 person the box in the middle says undefined? Where am I going wrong? I can see from the terminal that I am missing the LLM folder inside the ComfyUI models folder. Also there is an error saying naistyles.csv is not found?

Dauron Robinson

I have a local windows install, trying to use the image to video and getting the following error "torch.OutOfMemoryError: Allocation on device"

nando Pluister

I have a lot of missing nodes, but when I go to the missing node installer in the manager it says there are 0 to download.

Janne Kallio

My goal and idea was more like the fact that we will now have lot of people testing this and we should use/share the gained knowledge.

benoit corvol

It's working like a charm for me, but it could be awesome if you explain how to add some models to the auto install file ? thank's

Firmy Van Bas

I've tried re-installing multiple times, follow all the instructions and still get the same errors below: ReActorFaceBoost (In group node 'workflow>REACTORFACESWAP') workflow>REACTORFACESWAP ReActorFaceSwap (In group node 'workflow>REACTORFACESWAP') ReActorMaskHelper (In group node 'workflow>REACTORFACESWAP') workflow>faceswap2 workflow>faceswap3 PulidFluxInsightFaceLoader (In group node 'workflow>pluidlux') workflow>pluidlux PulidFluxEvaClipLoader (In group node 'workflow>pluidlux') PulidFluxModelLoader (In group node 'workflow>pluidlux') ApplyPulidFlux (In group node 'workflow>pluidlux') workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpluidlux

Firmy Van Bas

I don't see any errors while installing the InsightFace, once it's done, it shuts the cmd window down. Here is the errors in my cmd when I load up the workflow [] FETCH DATA from: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE] [ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node. Traceback (most recent call last): File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\server.py", line 564, in get_object_info out[x] = node_info(x) ^^^^^^^^^^^^ File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\server.py", line 531, in node_info info['input'] = obj_class.INPUT_TYPES() ^^^^^^^^^^^^^^^^^^^^^^^ File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in INPUT_TYPES "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 148, in "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "pathlib.py", line 931, in iterdir FileNotFoundError: [WinError 3] The system cannot find the path specified: 'Z:\\COMFY_UI\\ComfyUI_windows_portable\\ComfyUI\\models\\LLM' Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI

Ken L. Black

Everything is working great so far, so thanks! Is there a way to add other Flux checkpoints, which (I THINK), go into the checkpoints folder(?). The first prompt specifies a unet to add, rather than a checkpoint. One reason I ask is that my understanding is that the alternative checkpoints often have the VAE baked in, and as I mentioned, go in the checkpoints folder of the installation.

Two Magic Pixels

At Local Install, step 4, where is the file "V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json" located (where do we load it from)?

BladeOkelly

hi! thank you, thats a wondeful work!

McGenie

I am trying the Image to Video - ahad to download a few files manually. But now I get this error: DownloadAndLoadCogVideoModel 'CogVideoXTransformer3DModel' object has no attribute 'context_embedder' Any ideas as to what the issue is? The rest is really fun and working fine :-)

Aitrepreneur

the fact that the InsightFace installer closes automatically is not a good sign, that means that there was an error during install. Go inside the python embeded folder, open a cmd window and drag and drop the insightface installer inside then press enter and take a look at the errors (if there is one)

Aitrepreneur

Read the patreon post at the beginning man, every step is explained

Aitrepreneur

you ran out of vram, meaning your GPU didn't have enough vram to complete the operation, you can enable sysmem fallback in the nvidia control panel to avoid that issue in the future

Aitrepreneur

Read the patreon post at the beginning man, every step is explained

Aitrepreneur

just download them and put them in the models/unet folder

Aitrepreneur

you need to put them in the models/unet folder checkpoints is for SD models

Aitrepreneur

hmm that error usually means that the video model downloaded was corrupted for some reason. Go inside the ComfyUI_windows_portable\ComfyUI\models folder and delete the cogvideo folder. Then download the V2-MODELS-ULTIMATE_FLUX_INSTALL.bat file, put it inside the ComfyUI_windows_portable\ComfyUI folder and then run it, it will automatically install all the missing models/files

JB11C

Is there any way we could get a rough estimate for how large the entire package is? My install is over 50GB at the moment and still going, so it would be helpful to know how much we're installing so we can open up space if needed on the PC. tyty! <3

McGenie

Thanks - that worked.

Firmy Van Bas

Ok I did that, first it didn't want to install, but it said to update the pip, which I did, and then it installed insight face correctly, but now when I launch comfy UI, i get this and it doesn't load ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Cannot import Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced module for custom nodes: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Traceback (most recent call last): File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\nodes.py", line 2012, in load_custom_node module_spec.loader.exec_module(module) File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node\__init__.py", line 23, in from .nodes import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node\nodes.py", line 15, in from insightface.app.common import Face File "Z:\COMFY_UI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\insightface\__init__.py", line 16, in from . import model_zoo File "Z:\COMFY_UI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\insightface\model_zoo\__init__.py", line 1, in from .model_zoo import get_model File "Z:\COMFY_UI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\insightface\model_zoo\model_zoo.py", line 11, in from .arcface_onnx import * File "Z:\COMFY_UI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\insightface\model_zoo\arcface_onnx.py", line 10, in import onnx File "Z:\COMFY_UI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx\__init__.py", line 77, in from onnx.onnx_cpp2py_export import ONNX_ML ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Cannot import Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node module for custom nodes: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Z:\COMFY_UI\ComfyUI_windows_portable ############################################ Z:\COMFY_UI\ComfyUI_windows_portable\custom_nodes\ComfyUI-NAI-styler\CSV ############################################ [] ############################################ ------------------------------------------ Comfyroll Studio v1.76 : 175 Nodes Loaded ------------------------------------------ ** For changes, please see patch notes at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/blob/main/Patch_Notes.md ** For help, please see the wiki at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/wiki ------------------------------------------ [comfyui_controlnet_aux] | INFO -> Using ckpts path: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts [comfyui_controlnet_aux] | INFO -> Using symlinks: False [comfyui_controlnet_aux] | INFO -> Using ort providers: ['CUDAExecutionProvider', 'DirectMLExecutionProvider', 'OpenVINOExecutionProvider', 'ROCMExecutionProvider', 'CPUExecutionProvider', 'CoreMLExecutionProvider'] Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\node_wrappers\dwpose.py:26: UserWarning: DWPose: Onnxruntime not found or doesn't come with acceleration providers, switch to OpenCV with CPU device. DWPose might run very slowly warnings.warn("DWPose: Onnxruntime not found or doesn't come with acceleration providers, switch to OpenCV with CPU device. DWPose might run very slowly") [rgthree-comfy] Loaded 42 extraordinary nodes. 🎉 WAS Node Suite: OpenCV Python FFMPEG support is enabled WAS Node Suite Warning: `ffmpeg_bin_path` is not set in `Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui\was_suite_config.json` config file. Will attempt to use system ffmpeg binaries if available. WAS Node Suite: Finished. Loaded 218 nodes successfully. "Don't be pushed around by the fears in your mind. Be led by the dreams in your heart." - Roy T. Bennett Import times for custom nodes: 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-mxToolkit 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Logic 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Universal-Styler 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_JPS-Nodes 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-various 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Custom-Scripts 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Chibi-Nodes 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Frame-Interpolation 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_UltimateSDUpscale 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\Derfuu_ComfyUI_ModdedNodes 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\x-flux-comfyui 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_Comfyroll_CustomNodes 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-KJNodes 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-dream-project 0.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux 0.1 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Long-CLIP 0.1 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-VideoHelperSuite 0.1 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack 0.2 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2 0.2 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-tensorops 0.3 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager 0.3 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use 0.3 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper 0.8 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_LayerStyle 1.0 seconds: Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui 2.3 seconds (IMPORT FAILED): Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux 2.7 seconds (IMPORT FAILED): Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node 2.7 seconds (IMPORT FAILED): Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced Starting server Traceback (most recent call last): File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\main.py", line 261, in loop.run_until_complete(run(server, address=args.listen, port=args.port, verbose=not args.dont_print_server, call_on_start=call_on_start)) File "asyncio\base_events.py", line 654, in run_until_complete File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\main.py", line 166, in run await asyncio.gather(server.start_multi_address(addresses, call_on_start), server.publish_loop()) File "Z:\COMFY_UI\ComfyUI_windows_portable\ComfyUI\server.py", line 840, in start_multi_address await site.start() File "Z:\COMFY_UI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\aiohttp\web_runner.py", line 119, in start self._server = await loop.create_server( ^^^^^^^^^^^^^^^^^^^^^^^^^ File "asyncio\base_events.py", line 1536, in create_server OSError: [Errno 10048] error while attempting to bind on address ('127.0.0.1', 8188): only one usage of each socket address (protocol/network address/port) is normally permitted Z:\COMFY_UI\ComfyUI_windows_portable>pause Press any key to continue . . .

Randall Wainwright

I tried this, but I'm still getting these errors, any ideas? Here's the errors I see on startup during load: File "C:\ComfyUI\python_embeded\Lib\site-packages\onnx\__init__.py", line 77, in from onnx.onnx_cpp2py_export import ONNX_ML ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Cannot import C:\123abc\ComfyUI\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux module for custom nodes: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. 4.2 seconds (IMPORT FAILED): C:\ComfyUI\ComfyUI\custom_nodes\comfyui-reactor-node 4.3 seconds (IMPORT FAILED): C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced 4.6 seconds (IMPORT FAILED): C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux

Sam Shennan

oh yikes! I was doing this on a mobile hotspot and though it would be no more than 30!

James Hanggi

What would be super cool is if this model included a much higher res edition of instant id!!

Alex Mahr

I'm having a problem getting the image2video to work. i hit queue and it just crashes

Dauron Robinson

I enabled that from my control panel, restarted and still getting the same error 😞

benoit corvol

No I mean to add them in the auto install .sh or .bat so if I run the auto install file that will download them automatically but thanks for your answers

tc

I first attempted to update my existing ComfyUI install using the V2-MODELS-ULTIMATE_FLUX_INSTALL.bat method. That failed. So, I deleted my original install and freshly-installed ComfyUI using V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat After ComfyUI’s first launch, I installed all missing nodes and restarted. After several restarts and attempts to get all nodes to install, ComfyUI continues to display these errors: —————————— When loading the graph, the following node types were not found: FluxGuidance (In group node 'workflow/positive prompt guidance') workflow/positive prompt guidance workflow/Negative Prompt LoadFluxControlNet (In group node 'workflow/Flux Controlnet') workflow/Flux Controlnet ApplyFluxControlNet (In group node 'workflow/Flux Controlnet') XlabsSampler (In group node 'workflow/XLabs Sampler') workflow/XLabs Sampler ModelSamplingFlux (In group node 'workflow/Other Nodes') workflow/Other Nodes SetUnionControlNetType (In group node 'workflow/fff') workflow/fff workflow/outpaintfluxendworkflow/pulidtxt PulidFluxInsightFaceLoader (In group node 'workflow/pluidlux') workflow/pluidlux PulidFluxEvaClipLoader (In group node 'workflow/pluidlux') PulidFluxModelLoader (In group node 'workflow/pluidlux') ApplyPulidFlux (In group node 'workflow/pluidlux') workflow/Positive Prompt workflowsampler and seed upscaler workflowExtra nodes upscaler workflowFlorence workflowFlux Controlnet workflowImage gen sampler workflowOther Nodes workflowParam workflowsampler img workflowOther Nodes 2 workflowNegative Prompt workflowSampler 3 workflowNodes 3 DualCLIPLoaderGGUF UnetLoaderGGUF workflowXLabs Sampler workflowpositive prompt guidance workflowAdetailer workflowfff workflowoutpaintsize1 workflowoutpaintfluxend workflowoutpaintflor workflowVIDEOUPSCALER workflowvideoencodecode workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpullidpara workflowpullflo workflowpulidtxt workflowpluidlux Nodes that have failed to load will show as red on the graph.

Adam Booth

Ok I have a new error I cant resolve - sorry. When I go to create an image I get the following in the terminal which isnt much help in pining down the issue so hoping some one knows? got prompt Using pytorch attention in VAE Using pytorch attention in VAE G:\comfyUI\ComfyUI_windows_portable>pause Press any key to continue . . . After this it closes.

Adam Booth

Actually I fixed this, it was due to a lack of hard disk space.

Ken L. Black

Thank you! That is very good to know. I appreciate it! This workflow took a LOT of work on your part, and I appreciate it.

Einziger

WOW! I just tried the image to video, this is amazing! Thank you, you deserve every penny I pay for this patreon! Keep up the good work!

Chris Johnston

I'm running into this problem where I start your pod template on Runpod and it shows 5 ports to access. After restarting a stopped pod, it only comes back online with 2 urls to connect too, and 7777 (VScode) isn't there. any thoughts?

Sheen

Im getting this as well

Sheen

Ahh redragging in the workflow fixed this

FreerTech

New windows 10 installation...Local install....I run the bat file I chose 12-16 graphics card... I Chose to download the flux model, and it starts installing git, it starts the git installation, and then windows disappears. I did not get any more confirmation that it is still installing, and I also did not have the webui launch. Am I doing something wrong?

tc

Unfortunately, redragging the workflow into the browser doesn't work for me.

Chris D

nope its just a broke bat for whatever reason. same issue here. its just not working.

Aitrepreneur

well pulid or instantid are similar tools but yes instantid is not compatible with flux as of right now and not sure they will

Aitrepreneur

probably running out of resources, this is an intensive tool as I said in my video during recording my recording software crashed as well so yeah

Aitrepreneur

follow the patreon post, I explain everything in the beginning

Aitrepreneur

no idea, that's very strange indeed, maybe ask runpod support?

Aitrepreneur

it's not a broke bat, just open a new cmd window and drag and drop the installer inside then press enter this will avoid any crashing and will at least give us an error to work with. Make sure you don't have any antivirus or vpn running and that your internet connection is solid, there are like 80gb+ of files to dl

Bunny Swallows

This is an insane amount of work! i've gotten a few things to run with no issue or minimal fixes. But it seems that every time I need to use a model (Runpod version), they are never there. at least the flux versions. I'm guessing it never downloads? If that's the case, is there a way i can download it and add it to my session in some way? Maybe I'm making a mistake somewhere.

Cássio Braga

My ComfyUI is installed through Pinokio, and it doesnt have this "python_embeded" folder page so I cant run INSIGHTFACE_AUTO_INSTALL.bat inside it. I need to do this because when starting my ComfyUI with the ultimate workflow Im still receiving some nodes in red. Any ideas?

Dawan Star

I have an issue when I use FP8 PULID it say Prompt outputs failed validation UNETLoader: - Value not in list: unet_name: 'flux1-dev-fp8.safetensors' not in []

Slavik Riff

Hi! Thank you very much for your work! I have a problem with PULID workflow - when I execute the queue, I got this error: LongCLIPTextEncodeFlux 'visual.layer1.0.conv1.weight' That happens on both GGUf and FP8 groups, re-updated and re-instalded LongClip node, same result... any ideas? Thanks!

VitalClean

Erreur using your installer on runpod, when installing the missing nodes, ComfyUI Impact Pack install failed: Unknown Error please help

Carasibana

I was having issues with reactor and PuLID nodes not running due to onnx DLL errors; finally got it working by uninstalling onnx ( I bielieve it was 1.15.1 ) and reinstalling a specific version ( found suggestion at https://github.com/Gourieff/comfyui-reactor-node/issues/400#issuecomment-2305967670 ) python_embeded\python.exe -m pip uninstall onnx onnxruntime onnxruntime-gpu python_embeded\python.exe -m pip install onnx onnx==1.14.0 onnxruntime-gpu==1.15.0 Thought I'd share if anyone was having the same issue

FreerTech

I manually installed Git(the process that it kept hanging on) and its currently installing. Weird, It's a brand new install of windows.

Aitrepreneur

everything should be downloaded, it takes a lot of time though, follow the runpod step by step and check in with the logs command so that you know that everything is downloading, you might have to download missing custom nodes multiple times but it works

Aitrepreneur

Follow the patreon post, you haven't even loaded the json workflow

Aitrepreneur

my files and installers are made in mind for my own installers not third parties, also you'll be missing all models and files anyway if you use anything else

Aitrepreneur

is the 'flux1-dev-fp8.safetensors model correctly downloaded in the models/unet folder?

Aitrepreneur

never saw this error before, send me more info in dm + screenshots of everything

Aitrepreneur

follow the patreon post and check in with the logs command so that you know that everything is downloading, you might have to download missing custom nodes multiple times. You need to let everything download step by step before proceeding and it takes a while, if you see an error, just uninstall that particular node, restart, wait for everything to work again, refresh the page, then reinstall all the missing nodes again and be patient

Aitrepreneur

the INSIGHTFACE_AUTO_INSTALL.bat should already do that, as well as the CUDA_PATH_ONNXRUNTIME_ERROR_FIXER.bat file

Aitrepreneur

that's because everything is written before: !!PLEASE READ THE POST BEFORE DOING THE INSTALL!! ALSO NOTE: After the missing nodes download, a white screen will appear etc...etc... it's right there man, very first sentence

Ricardo Campos

now I can see it! i did not, so strange... thanks!!

Ricardo Campos

Once everything is loaded... when I open in Runpod the port 3000... the manager menu doesn´t exist... any suggestion? Thanks!

Carasibana

I hadn't tried the cuda path fixer, but did try the insightface auto install bat but had the problem persist. Regardless its working now :)

Ramon Rodgers

So I started a completely new install in a fresh Anaconda environment. This time with nothing, not even version one or models. I got the same result, It just keeps installing the same 6 files in the "\models\facerestore_models" folder. --- Is there a way to just get it to stop trying to update the install so ComfyUI will launch? T have all 6 files backed up so I dont really need it to check anymore.

Aitrepreneur

are you using my template? The manager is already pre installed with the template

Aitrepreneur

fresh anaconda env? Why would you do that when it's a portable version?? Just use my V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat and launch it, that's it

Ramon Rodgers

In case I corrupted some download the first time. I was just trying anything i could think of and thought to just start over. You bat file installs everything beautifully. The the very last step is to install the Face Restoration models. And it perpetually thinks the files are not there and continually and repeatedly re-downloads them... I have the files so I just need it to recognize that

Stephen Murray

I use the template for the install but I am missing nodes that are not available for download

Cesar Gaytán

Thanks, you were right, I just had to drag and drop the V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json on the window and then select "install missing nodes" on the manager window

Cesar Gaytán

For those having issues installing the script in windows, I'll write a checklist below. 1.- Download "V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat" 2.- Run CMD and go to your folder with the bat file 3.- run the bat file (important the instalation needs to be on C: Disk that was my first issue) 4.- once finished go to "manager" and click on "update all" 5.- go to "install missing custom nodes" and if nothing is able to install click close and then close again to go to the main window. 6.- download "V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json" and drag and drop the file in to your browser window where comfyUI is 7.- go to "manager" then "install missing custom nodes" and finally install the missing nodes 8.- Restart comfyUI and enjoy ! Now almost everything is working for me, I just have to solve why image to video is not working for me. I'll update on that later My current error for just " image to video" is this: https://ibb.co/9Y65R1G Any help will be appreciate it =)

Cesar Gaytán

Now almost everything is working for me, I just have to solve why image to video is not working My current error for just " image to video" is this: https://ibb.co/9Y65R1G Any help will be appreciate it =) For those having issues installing the script in windows, I'll write a checklist below. 1.- Download "V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat" 2.- Run CMD and go to your folder with the bat file 3.- run the bat file (important the instalation needs to be on C: Disk that was my first issue) 4.- once finished go to "manager" and click on "update all" 5.- go to "install missing custom nodes" and if nothing is able to install click close and then close again to go to the main window. 6.- download "V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json" and drag and drop the file in to your browser window where comfyUI is 7.- go to "manager" then "install missing custom nodes" and finally install the missing nodes 8.- Restart comfyUI and enjoy !

Cássio Braga

same problem here with the image to video

MonoChromancer

I managed to fix almost all of my errors and tested many areas so far and they work (I had the onnx error and had to run the comfy update etc). However the video generation halts on a weird error I don't see above: Failed to validate prompt for output 1035: * CLIPLoader 1053: - Value not in list: clip_name: 't5xxl_fp8_e4m3fn.safetensors' not in ['ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors', 't5xxl_fp16.safetensors'] Output will be ignored Failed to validate prompt for output 1047: * (prompt): - Return type mismatch between linked nodes: images, CLIP != IMAGE * VHS_VideoCombine 1047: - Return type mismatch between linked nodes: images, CLIP != IMAGE Output will be ignored Prompt executed in 0.01 seconds got prompt Failed to validate prompt for output 1035: * DF_Get_image_size 1062:2: - Required input is missing: image * ImageScale 1062:6: - Required input is missing: image * ImageUpscaleWithModel 1062:1: - Required input is missing: image Output will be ignored Failed to validate prompt for output 1047: * (prompt): - Return type mismatch between linked nodes: images, CLIP != IMAGE * VHS_VideoCombine 1047: - Return type mismatch between linked nodes: images, CLIP != IMAGE Output will be ignored Prompt executed in 0.07 seconds Watching your video closely I think the videoencodecode node or the Video Combine 🎥 node may have changed. I downloaded and loaded the workflow directly from the site here and looking at the nodes in my flow vs your video (its hard to tell) but it looks like there is a "clip" and "images" outflow from videoencodecode, and now "clip" connects to Video Combine 🎥 on the "images" inflow, but you cant swap the outflow connection to the "images outflow" and the 3 outflows from "images" are not by default connected to the 3 inputs on the VIDEOUPSCALER node. I can only partially reconnect these to the similar setup you have but this is what I am seeing by default and I can't seem to easily restore it. I would post an image but it doesn't allow that in the comments This was the thing I was most excited to get working with and I am not sure how to proceed. thanks for your time

FreerTech

I keep getting Reactor, Easy Use and Pulid, node import failures. I try to fix, and the repair fails, I clear the workflow and re add it and it still fails. I run the INSIGHTFACE_AUTO_INSTALL.bat CUDA_PATH_ONNXRUNTIME_ERROR_FIXER.bat and it still fails after I restart and uninstall and re-install the plugins. I also get the following errors/info when starting(I removed anything that wasn't an error or system info): FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\Tom\\ComfyUI_windows_portable\\ComfyUI\\custom_nodes\\ComfyUI-Easy-Use\\wildcards\\example.txt' Cannot import C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use module for custom nodes: [Errno 2] No such file or directory: 'C:\\Users\\Tom\\ComfyUI_windows_portable\\ComfyUI\\custom_nodes\\ComfyUI-Easy-Use\\wildcards\\example.txt' [Impact Pack] Wildcards loading done. Total VRAM 12282 MB, total RAM 32581 MB pytorch version: 2.3.1+cu121 Set vram state to: NORMAL_VRAM Device: cuda:0 NVIDIA GeForce RTX 4070 SUPER : cudaMallocAsync ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Cannot import C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux module for custom nodes: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. File "C:\Users\Tom\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx\__init__.py", line 77, in from onnx.onnx_cpp2py_export import ONNX_ML ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Cannot import C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced module for custom nodes: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\node_wrappers\dwpose.py:26: UserWarning: DWPose: Onnxruntime not found or doesn't come with acceleration providers, switch to OpenCV with CPU device. DWPose might run very slowly warnings.warn("DWPose: Onnxruntime not found or doesn't come with acceleration providers, switch to OpenCV with CPU device. DWPose might run very slowly") ←[34mWAS Node Suite ←[93mWarning: ←[0m`ffmpeg_bin_path` is not set in `C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui\was_suite_config.json` config file. Will attempt to use system ffmpeg binaries if available.←[0m 0.2 seconds (IMPORT FAILED): C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use 1.4 seconds (IMPORT FAILED): C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced 1.4 seconds (IMPORT FAILED): C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux 1.4 seconds (IMPORT FAILED): C:\Users\Tom\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node File "pathlib.py", line 931, in iterdir FileNotFoundError: [WinError 3] The system cannot find the path specified: 'C:\\Users\\Tom\ComfyUI_windows_portable\\ComfyUI\\models\\LLM' Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: C:\Users\Tom\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: C:\Users\Tom\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: C:\Users\Tom\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: C:\Users\Tom\ComfyUI_windows_portable\ComfyUI [ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node.

Ricardo Campos

Yes, it was the disk size config. I was more than 1 hour loading and configurating, but always nodes and LLM error... it is not affordable to use it in runpod. It´s a petty :( so sad

Firmy Van Bas

oh whoa, I've had lots of issues and I've been installing it on my Z: drive. Why does it have to be on C: ? Obviously this looks like why Im having issues, but I can't really install it on C:

Cesar Gaytán

I see theres like a visual line (link) broken from image to image, if you compare my install with the video but dont know how to fix it.

Cesar Gaytán

Maybe the scripts have hardcoded c: i just tried to installed the system as similar as in the video, and worked, i did the install on c: because the errores where it couldn’t find files.

tc

Thanks, K. Even though I was working with what I thought was a most-recent ComfyUI when I followed the “IF YOU ALREADY HAVE COMFYUI INSTALLED” instructions, and even though I could have sworn I made sure ComfyUI was updated after my first subsequent failed fresh install, the critical step to get things working for me was in your “LOCAL INSTALL” Step 4: click "Update comfyui". Again, I could have sworn I updated cUI both with my existing install and the subsequent first fresh one. Using your .bat file installer, is there a possibility that the install could fail if you “Install Missing Custom Nodes” BEFORE updating ComfyUI ... and that if the UI isn’t updated FIRST that any future attempts to do so will not fix a compromised install created by installing custom nodes first? THANKS for all you do! Now off to play ...

philmutton

Thank you. This worked perfectly.

Starkselle

Can this workflow be used with other checkpoints than FLUX? I'm used to SDXL and all the Loras I use are SDXL. Thanks

Dylan Barnett

I got the image to video to work once but now when I click queue prompt it seems to stop with the video combine panel image input with a red circle. Edit: never mind. I right-clicked on that node and chose "repair node" and it fixed it. Fantastic job putting this all together. Thank you!

Randall Wainwright

This is the way. INSIGHTFACE_AUTO_INSTALL.bat always installs onnxruntime-gpu==1.15.1 which has a conflict, but this works: python_embeded\python.exe -m pip install onnx onnx==1.14.0 onnxruntime-gpu==1.15.0

Markus

Same problem here

Lega Lexo

Hello, i cant get image to video woroking because the videoupscaler has no inputs for image, DG_Get_image_size image and ImageScale image even when loading the fresh copy of the workflow downloaded here please help idk where to connect them

Aitrepreneur

There was an update that broke image to video, I just updated the workflow again, it should work fine now, you need to redownload the workflow from the other post again

Aitrepreneur

it works fine, I just did it like an hour ago, you just need to be patient and follow what the log says, I might have to do like an exclusive video for patreons maybe showing step by step slowly how to do that. But you just need to follow everything step by step, update, install missing nodes, restart and refresh the webui, then there will still be some left missing nodes to install, restart again and once all the missing nodes are installed, reload the workflow again.

Aitrepreneur

There was an update that broke image to video, I just updated the workflow again, it should work fine now, you need to redownload the workflow from the other post again

Aitrepreneur

not sure where the initial issue is since it seems like a bunch of stuff is missing from the start, just use the installer and reinstall from scratch in a new folder

Aitrepreneur

There was an update that broke image to video, I just updated the workflow again, it should work fine now, you need to redownload the workflow from the other post again

Aitrepreneur

There was an update that broke image to video, I just updated the workflow again, it should work fine now, you need to redownload the workflow from the other post again

MonoChromancer

That’s what I suspected, thank you so much for all your amazing work!

Keres666

Hey using Runpod, I had an issue where after getting the missing nodes and restart/refresh it was still telling me it was missing a bunch of things (workflowsomething something about 12 times) and the flow didn't work. I fixed my issue and just posting my fix here if anyone has that same problem. I just reloaded the json into comfyui again and everything seems to work...

MonoChromancer

Still getting: Failed to validate prompt for output 1191: * CLIPLoader 1210: the connections are correct in the new workflow but there is still an error, trying the C: drive trick to see if that fixes it

MonoChromancer

Moved to C: drive which seemed to work for a few people then did repair node which was something someone else tried and it worked but now I get this: CogVideoTextEncode 'SD1Tokenizer' object has no attribute 't5xxl' # ComfyUI Error Report ## Error Details - **Node Type:** CogVideoTextEncode - **Exception Type:** AttributeError - **Exception Message:** 'SD1Tokenizer' object has no attribute 't5xxl' ## Stack Trace ``` File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 323, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 198, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 169, in _map_node_over_list process_inputs(input_dict, i) File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 158, in process_inputs results.append(getattr(obj, func)(**inputs)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper\nodes.py", line 326, in process clip.tokenizer.t5xxl.pad_to_max_length = True ^^^^^^^^^^^^^^^^^^^^ ``` ## System Information - **ComfyUI Version:** v0.2.7-6-g2865f91 - **Arguments:** ComfyUI\main.py --windows-standalone-build - **OS:** nt - **Python Version:** 3.11.8 (tags/v3.11.8:db85d51, Feb 6 2024, 22:03:32) [MSC v.1937 64 bit (AMD64)] - **Embedded Python:** true - **PyTorch Version:** 2.5.1+cu121 ## Devices - **Name:** cuda:0 NVIDIA GeForce RTX 4080 : cudaMallocAsync - **Type:** cuda - **VRAM Total:** 17170956288 - **VRAM Free:** 15720078592 - **Torch VRAM Total:** 33554432 - **Torch VRAM Free:** 33381632 ## Logs ``` 2024-11-09 17:56:05,992 - root - INFO - Total VRAM 16376 MB, total RAM 65277 MB 2024-11-09 17:56:05,993 - root - INFO - pytorch version: 2.5.1+cu121 2024-11-09 17:56:05,994 - root - INFO - Set vram state to: NORMAL_VRAM 2024-11-09 17:56:05,994 - root - INFO - Device: cuda:0 NVIDIA GeForce RTX 4080 : cudaMallocAsync 2024-11-09 17:56:06,983 - root - INFO - Using pytorch cross attention 2024-11-09 17:56:07,914 - root - INFO - [Prompt Server] web root: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\web 2024-11-09 17:56:08,910 - ComfyUI-CogVideoXWrapper.custom_cogvideox_transformer_3d - INFO - sageattn not found, using sdpa 2024-11-09 17:56:08,919 - ComfyUI-CogVideoXWrapper.cogvideox_fun.transformer_3d - INFO - sageattn not found, using sdpa 2024-11-09 17:56:08,921 - ComfyUI-CogVideoXWrapper.cogvideox_fun.fun_pab_transformer_3d - INFO - sageattn not found, using sdpa 2024-11-09 17:56:09,640 - root - INFO - Total VRAM 16376 MB, total RAM 65277 MB 2024-11-09 17:56:09,640 - root - INFO - pytorch version: 2.5.1+cu121 2024-11-09 17:56:09,641 - root - INFO - Set vram state to: NORMAL_VRAM 2024-11-09 17:56:09,642 - root - INFO - Device: cuda:0 NVIDIA GeForce RTX 4080 : cudaMallocAsync 2024-11-09 17:56:13,736 - root - INFO - Import times for custom nodes: 2024-11-09 17:56:13,736 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Logic 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Universal-Styler 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-mxToolkit 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_JPS-Nodes 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Custom-Scripts 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_UltimateSDUpscale 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-various 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Chibi-Nodes 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Frame-Interpolation 2024-11-09 17:56:13,737 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\Derfuu_ComfyUI_ModdedNodes 2024-11-09 17:56:13,738 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\x-flux-comfyui 2024-11-09 17:56:13,738 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy 2024-11-09 17:56:13,738 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-KJNodes 2024-11-09 17:56:13,738 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_Comfyroll_CustomNodes 2024-11-09 17:56:13,738 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-dream-project 2024-11-09 17:56:13,738 - root - INFO - 0.0 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux 2024-11-09 17:56:13,738 - root - INFO - 0.1 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Long-CLIP 2024-11-09 17:56:13,738 - root - INFO - 0.1 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-reactor-node 2024-11-09 17:56:13,738 - root - INFO - 0.1 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-VideoHelperSuite 2024-11-09 17:56:13,738 - root - INFO - 0.2 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2 2024-11-09 17:56:13,739 - root - INFO - 0.2 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack 2024-11-09 17:56:13,739 - root - INFO - 0.3 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager 2024-11-09 17:56:13,739 - root - INFO - 0.3 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use 2024-11-09 17:56:13,739 - root - INFO - 0.3 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-tensorops 2024-11-09 17:56:13,739 - root - INFO - 0.3 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper 2024-11-09 17:56:13,739 - root - INFO - 0.4 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_LayerStyle 2024-11-09 17:56:13,740 - root - INFO - 1.1 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced 2024-11-09 17:56:13,740 - root - INFO - 1.6 seconds: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui 2024-11-09 17:56:13,740 - root - INFO - 2024-11-09 17:56:13,747 - root - INFO - Starting server 2024-11-09 17:56:13,748 - root - INFO - To see the GUI go to: http://127.0.0.1:8188 2024-11-09 17:59:50,231 - root - INFO - got prompt 2024-11-09 17:59:50,423 - root - ERROR - Failed to validate prompt for output 1191: 2024-11-09 17:59:50,424 - root - ERROR - * CLIPLoader 1210: 2024-11-09 17:59:50,424 - root - ERROR - - Value not in list: clip_name: 't5xxl_fp8_e4m3fn.safetensors' not in ['ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors', 't5xxl_fp16.safetensors'] 2024-11-09 17:59:50,424 - root - ERROR - Output will be ignored 2024-11-09 17:59:50,515 - root - ERROR - Failed to validate prompt for output 1187: 2024-11-09 17:59:50,515 - root - ERROR - Output will be ignored 2024-11-09 17:59:50,779 - root - INFO - Prompt executed in 0.25 seconds 2024-11-09 18:00:37,919 - root - INFO - got prompt 2024-11-09 18:00:38,079 - ComfyUI-CogVideoXWrapper.utils - INFO - Downloading model to: C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\models\CogVideo\CogVideoX-5b-I2V 2024-11-09 18:03:10,643 - ComfyUI-CogVideoXWrapper.utils - INFO - Encoded latents shape: torch.Size([1, 1, 16, 60, 90]) 2024-11-09 18:03:11,405 - root - ERROR - !!! Exception during processing !!! 'SD1Tokenizer' object has no attribute 't5xxl' 2024-11-09 18:03:11,416 - root - ERROR - Traceback (most recent call last): File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 323, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 198, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 169, in _map_node_over_list process_inputs(input_dict, i) File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\execution.py", line 158, in process_inputs results.append(getattr(obj, func)(**inputs)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\OldFlux Sept2024\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper\nodes.py", line 326, in process clip.tokenizer.t5xxl.pad_to_max_length = True ^^^^^^^^^^^^^^^^^^^^ AttributeError: 'SD1Tokenizer' object has no attribute 't5xxl' 2024-11-09 18:03:11,418 - root - INFO - Prompt executed in 153.39 seconds ``` ## Attached Workflow Please make sure that workflow does not contain any sensitive information such as API keys or passwords. ``` Workflow too large. Please manually upload the workflow from local file system. ``` ## Additional Context (Please add any additional context or steps to reproduce the error here)

Cesar Gaytán

Aitrepreneur just reply he fixed the WORKFLOW.json file, just redownload and import on COMFYUI by drag and drop it.

Ultimate Big Bad

I admit that i am new to this. Cannot execute because a node is missing the class_type property.: Node ID '#391' I keep getting this error and I don't know how to fix it. please help.

michael morris

without laughing at me... How do you start it once it's installed?

Tigon

can you put just the workflow json file somewhere? shouldnt need a .bat for that

Tigon

just kidding, i found it. im blind apparently

SlyNine

Which one did you install?

SlyNine

Did a local install, After all was said and done I could not get missing nodes to load.. But then I cleared the workflow and reloaded it and it solved the issue. Perhaps a step you might add in your guide?

Blue

When and where do I install this? V2-MODELS-ULTIMATE_FLUX_INSTALL.bat

Neal Barnett

Thanks for this comment. I was having the same issue and this fixed it.

aprisma

Hello there, thanx for your work. I totally followed your docs and also fixed the problem with naistyler_nodes.py --> for all that have a problem too rename in the file they entry "ComfyUI-NAI-styler" to "ComfyUI-Universal-Styler" very funny thing. If your workflow is still "in memory" so already loaded if comfyui starts i get an error in the console and a lot of nodes are red. But if I load the worflow afterwards i get the same error messages but your nodes working. Strange but ok :-) These are the error messages. Just for info if you have an idea. FETCH DATA from: C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE] [ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node. Traceback (most recent call last): File "C:\AI\ComfyUI_windows_portable\ComfyUI\server.py", line 564, in get_object_info out[x] = node_info(x) ^^^^^^^^^^^^ File "C:\AI\ComfyUI_windows_portable\ComfyUI\server.py", line 531, in node_info info['input'] = obj_class.INPUT_TYPES() ^^^^^^^^^^^^^^^^^^^^^^^ File "C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 150, in INPUT_TYPES "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 150, in "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "pathlib.py", line 931, in iterdir FileNotFoundError: [WinError 3] Das System kann den angegebenen Pfad nicht finden: 'C:\\AI\\ComfyUI_windows_portable\\ComfyUI\\models\\LLM' []

Christopher Carlson

Create a folder on your drive where you want this to be installed, such as C:\ComfyUI and put the file into that folder and double click it.

Christopher Carlson

Two steps forward and now one step back. I contacted you last week about problems installing and missing nodes. I followed your instructions to install INSIGHTFACE_AUTO_INSTALL and after doing so I was able to install missing nodes and that dialogue box is now empty. I now have a new problem, when I start ComfyUI, I get the following warning https://www.imghippo.com/i/hHtu6692W.png that tells me: Missing Node Types When loading the graph, the following node types were not found - workflowsampler and seed upscaler workflowExtra nodes upscaler workflowFlorence workflowFlux Controlnet workflowImage gen sampler workflowOther Nodes workflowParam workflowsampler img workflowOther Nodes 2 workflowNegative Prompt workflowSampler 3 workflowNodes 3 workflowXLabs Sampler workflowpositive prompt guidance workflowAdetailer workflowfff workflowoutpaintsize1 workflowoutpaintfluxend workflowoutpaintflor workflowVIDEOUPSCALER workflowvideoencodecode workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpullidpara workflowpullflo workflowpulidtxt workflowpluidlux But when I check for missing nodes, there none. If I try to ignore the message and just use the workflow, you can see form this picture, many nodes are red and unavailable to use https://www.imghippo.com/i/PRr2555qc.png What do I do now?

aprisma

I found out its important to start with no workflow in cache and do a clear before starting comfyui new and then pull the workflow in. Just as a hint.

Salvador Robles

Hi, this version is awesome but I got a error:

Christopher Carlson

Now I see I have the following errors when trying to load: [ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node. Traceback (most recent call last): File "D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI\server.py", line 564, in get_object_info out[x] = node_info(x) ^^^^^^^^^^^^ File "D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI\server.py", line 531, in node_info info['input'] = obj_class.INPUT_TYPES() ^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 150, in INPUT_TYPES "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Florence2\nodes.py", line 150, in "model": ([item.name for item in Path(folder_paths.models_dir, "LLM").iterdir() if item.is_dir()], {"tooltip": "models are expected to be in Comfyui/models/LLM folder"}), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "pathlib.py", line 931, in iterdir FileNotFoundError: [WinError 3] The system cannot find the path specified: 'D:\\ComfyUI Ultimate\\ComfyUI_windows_portable\\ComfyUI\\models\\LLM' Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\ComfyUI Ultimate\ComfyUI_windows_portable\ComfyUI

Blue

After I installed V2-MODELS-ULTIMATE_FLUX_INSTALL.bat there is a folder named ComfyUI. Can I put it in there?

Aitrepreneur

Guys come on read the patreon post eveything is written there

Hrothgar's Hoard

Grrrr.....getting a whole host of missing node types....ReactorFaceBoost, FeActorFaceSwap, ReactorMaskhelper, Int, workflow>outpainsize, JWFloatDiv, Mul, Sub, Add, LayerUtility:ImageBlendAdvance V2, easy imageSize, Workflow>outpaintfluxend, faceswap2, faceswap3, LongCLIPTextEncodeFlux, pulidtxt, TextC Concatenate, DF_GET_image_size, vidiupsl, ....the list goes on another 20 things. if I click X and continue to the FP8 Image Generation and click on Queue, I get"Error while deserializing header: MetadataIncompleteBuffer" No clue where to even begin.

Aitrepreneur

you might have changed or deleted a node, try re drag and dropping the workflow file inside the ui and try again

Aitrepreneur

click on the run_nvidia_gpu.bat file inside the ComfyUI_windows_portable folder

Aitrepreneur

Everything is written in the patreon post man, step by step

Aitrepreneur

that error is ok, it should not influence the workflow

Aitrepreneur

Everything is written in the patreon post man, step by step

aprisma

Thats true as long I load the workflow after. Great work. But yes we should read the docs very good :-)

Firmy Van Bas

Ok so continuation of my previous post. I have now installed it on the C: drive, as I've seen people saying it had to be (I was installing on another drive), but still get the same Reactor error, even though I installed InsightFace correctly and it said Succesfull install. I launched the ComfyUI update from the update folder and I am getting this error below. Can this be related to why it's not working, and how can I change those? ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. flet 0.24.1 requires packaging<24.0,>=23.1, but you have packaging 24.2 which is incompatible. inference-gpu 0.26.0 requires pillow<11.0, but you have pillow 11.0.0 which is incompatible. mediapipe 0.10.18 requires protobuf<5,>=4.25.3, but you have protobuf 5.28.3 which is incompatible. transparent-background 1.3.3 requires albucore==0.0.16, but you have albucore 0.0.20 which is incompatible.

Firmy Van Bas

Getting this error when launched ComfyUI File "C:\comfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx\__init__.py", line 77, in from onnx.onnx_cpp2py_export import ONNX_ML ImportError: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. Cannot import C:\comfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux module for custom nodes: DLL load failed while importing onnx_cpp2py_export: A dynamic link library (DLL) initialization routine failed. [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json

James Hanggi

Hello and thank you for another wild setup. In trying to run FP8 Pulid. I got the following error message:

James Hanggi

Expected Scalar Type Half But Found BFloat16

Part Time Artist

Can I create one pod for both the newer OMNI install as well as this, or should I create separate ones?

Hrothgar's Hoard

Redid the install. It comes up OK now. However, When I use GGUF Image Generation, I get "Prompt outputs failed validation UnetLoaderGGUF: - Value not in list: unet_name: 'flux1-dev-Q8_0.gguf' not in [] DualCLIPLoaderGGUF: - Value not in list: clip_name2: 't5-v1_1-xxl-encoder-Q8_0.gguf' not in ['ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors', 'longclip-L.pt', 't5xxl_fp16.safetensors', 't5xxl_fp8_e4m3fn.safetensors']" If I use FP8 Image Generation, I get "Error while deserializing header: MetadataIncompleteBuffer" Looking at https://github.com/comfyanonymous/ComfyUI/issues/3225 and it suggests that I may need to redownload the safetensors file. Does that sound right and if so, how do I do that?

Ron Harris-White II

I am using the Faceswap Workflow, and I get a ReActorMaskHelper error. It mentions "Could not run 'torchvision::nms' with arguments from the 'CUDA' backend." I ran the INSIGHTFACE_AUTO_INSTALL.bat, but I get the same error. Any ideas?

dattrax

I'd like to get the installer up and running, but it looks like the local install is for windows only and the runpod is for remote. If I want it to work via WSL and a ubuntu install then I guess its a manual process?

Aitrepreneur

You could but better to create a separate one since you'll be using different templates anyway

Aitrepreneur

which option did you choose when installing? What Q model version was installed? It might say this just because you have chosen either option 1 or 2 and you just need to select the right model in the model selection node

Aitrepreneur

have you tried the CUDA_PATH_ONNXRUNTIME_ERROR_FIXER.bat file as well?

Aitrepreneur

I don't use linux or mac so I cannot support either since I can't make sure that they work correctly. Altough I'm pretty sure that the runpod installer can be easily modified to work for linux though, you might just need to manuall download and extract the portable comfyui version and put that installer inside the comfui portable/comfyui folder and run it, might work just like that

dattrax

I'm just using the windows script as a template at the moment and converting to linux. I'm fairly sure it'll work

Ivan P.

I am trying to run this workflow on service such as mimic pc and other virtual gpuAny other setting that I should do differently?. but run into serveral issues. Anyway you could help me?

Cesar Gaytán

When I try to use FP8 image to prompt I'm getting the error: "DownloadAndLoadFlorence2Model, Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory C:\V2-ULTIMATE_FLUX-COMFYUI-MANAGER\ComfyUI_windows_portable\ComfyUI\models\LLM\Florence-2-large." but if I go to manager I see I already have installed: ComfyUI-Florence-2, ComfyUI-Florence2. Any idea of what's wrong? Heres an image of the error: https://ibb.co/cDG2nbG

Cesar Gaytán

Hi, when I'm using "image to video", the video with 8 frames looks great but the one that generates after that one with 24 frames, always is low res, as you can see on the image, What am I doing wrong? Image showing the issue: https://ibb.co/LYznQhJ Thanks

Averagenova

Ran the V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat. No webUI came up. Went into ComfyUI-Manager folder and there is no exe there. How does this thing start ? Must be a me problem but got super annoyed . LoL. This is literally the first step.

GlobalGrooveWORLD

I can't understand from where to download this file??????. “2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json.” Why make it so that nothing is clear????? Well, there are attached files three bat and one sh, but WHERE IS THIS FILE????????

Markus

Click on "show more" and read the text. The link is there.

Matteo Di Lucca

Hey K, a quick question: 1st time installing comfy, do i need to install python before or the auto-installer will actually do that? Thanks!

Hrothgar's Hoard

I selected 3; I have a 3090 card. The model selected in unet loader(GGUF) is flux1-dev-Q8_0.gguf....ummm it was. I just clicked on the model and it now says "undefined" and doesn't give me another choice anymore.

Hrothgar's Hoard

I looked for .gguf files and found flux1-dev-fp8.safetensors, flux1-dev-Q8_0.gguf, and flux1-schnell-Q8_0.gguf in the models/unet folder.

GlobalGrooveWORLD

I don't know where exactly to look for this? I've already installed everything through the batnik, I've installed the missing nodes in the comfi itself and I got these errors when loading.

Ryan Clancy

It's on a different post. I have to agree. It would have been nice to see it linked here again. You can find the post a few posts back.

MonoChromancer

I have the exact same problem, 8 frame is fine, higher Rez looks corrupted and terrible. Lowering the upscale resolution factor seems to help a little but not much. I’ll post here if I find a solution but considering how powerful and amazing this tool is I hope we can collectively find a solution because it makes the in between frames and full resolution not possible as the video is unusably corrupted

Aitrepreneur

not sure, never used those before, what issues?

Aitrepreneur

try the GGUF one, this error actually means that you are running out of memory for the operation

Aitrepreneur

decrease the crf value in the last node to 10

Aitrepreneur

there must have been an issue during install. Reinstall again in a new folder and check the log for an error if there is one

Aitrepreneur

you can run the update_comfyui_and_python_dependencies.bat file present in the ComfyUI_windows_portable\update folder for the update

Aitrepreneur

Everything is listed and linked to this post though

Gerard Colin

Im getting an odd error when running the PULID workflow: SamplerCustomAdvanced forward_orig() takes from 7 to 9 positional arguments but 10 were given Everything else work and I do not believe I am missing any nodes or models Thoughts?

Cesar Gaytán

I did and got the same errors here you can see the errors: https://ibb.co/Pxws6k0 And I dont thing is a ram or vram issue, I have: - Intel(R) Core(TM) i9-10900F CPU @ 2.80GHz 2.81 GHz - 64gb RAM - GPU: TUF-RTX3090-O24G-GAMING with 24GB VRAM What else could be? Thanks

Cesar Gaytán

I can't find any solution, let's see if Aitrepreneur help us a little :)

Yukon Cornelius

Did you get this sorted out? I am getting the exact same error and I think it has to do with the #1162 pulidtxt box. I am pretty sure it doesnt have the right clip installed. it said something about longclip at first but now i dont see that as an option. downloading (BeichenZhang/LongCLIP-L) right now and gonna see if that works when I toss it in the clip model folder. Edit: Yeah that fixed it. I think the problem is just that some things automatically download when they are needed for the first time, and other things do not. Which is understandable, this is a massive workflow with a lot of moving parts. so basically just download (BeichenZhang/LongCLIP-L) and toss it in the model/clip folder, and then it will probably get almsot all the way done and then hang up and automatically download the (pulid_flux_v0.9.0.safetensors) file, which is a little over a gig so give it a sec. but now I am getting an error (forward_orig() takes from 7 to 9 positional arguments but 10 were given) at the very end it seems so... off to figure that out now, lol.

Yukon Cornelius

I am getting the same error. did you find any solutions? I get: SamplerCustomAdvanced: forward_orig() takes from 7 to 9 positional arguments but 10 were given and on the comfy ui banner error message it says : TypeError 1164:6 SamplerCustomAdvanced EDIT: I actually am also getting the error on the fp8 image generation. Which I had not even messed with really. "TypeError 399:3 SamplerCustomAdvanced" "SamplerCustomAdvanced: forward_orig() takes from 7 to 9 positional arguments but 10 were given"

VXLM

I read that its a backwards compitability issue after the latest pulid update - I am just not smart enough to fix it :D

Yukon Cornelius

Well I hope we can find a solution soon. This Smut isn't going to make itself!

Blue

When I use the upscaler it just stays at 0% when loading the model. Any idea what that could be? I've tried it on two different installs. (I didn't touch anything) Edit: Upscaler working on option 3

Gerard Colin

No solution yet. I tried a different workflow but the problems seems to lie with PULID itself as I am getting the same error.

Math

think same problem here, i have : (IMPORT FAILED) ReActor Node for ComfyUI and (IMPORT FAILED) ComfyUI-PuLID-Flux-Enhanced in install missing custom nodes

Death_Wish

I did the all in one runpod install and it acts like the checkpoints or vae never were installed Prompt outputs failed validation DualCLIPLoader: - Value not in list: clip_name2: 't5xxl_fp16.safetensors' not in [] - Value not in list: clip_name1: 'ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors' not in [] UNETLoader: - Value not in list: unet_name: 'flux1-dev-fp8.safetensors' not in [] VAELoader: - Value not in list: vae_name: 'ae.safetensors' not in ['sdxl_vae.safetensors']

Gerard Colin

I had that issue as well but I resolved it with brute force. I reinstalled comfy UI and installed from scratch. Not elegant but it worked. Still have the other issue though

Judi Godvliet

Hi AIOverlord. I tried to install 3 times the automatic installer for Flux & ComfyUI (v2) from scratch. After updating everything i get still the message below. I don't see any missing custom nodes in the addon manager, any idea how to solve this?: Missing Node Tpes: When loading the graph, the following node types were not found workflowsampler and seed upscaler workflowExtra nodes upscaler workflowFlorence workflowFlux Controlnet workflowParam workflowsampler img workflowOther Nodes 2 workflowNegative Prompt workflowSampler 3 workflowNodes 3 workflowOther Nodes workflowImage gen sampler workflowXLabs Sampler workflowpositive prompt guidance workflowAdetailer workflowfff workflowoutpaintsize1 workflowoutpaintfluxend workflowoutpaintflor workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpullidpara workflowpullflo workflowpulidtxt workflowpluidlux workflowvid encod workflowvidiupsl

Ron Harris-White II

Is there a method in your Uber ComfyUI Workflow where it will save EXIF metadata into the generated images?

LW

I also have this problem. Missing nodes but Manager says that there are no missing nodes to install. I went through the list of all nodes and there were a few that offered the option to "try to install", a reinstall, I suppose. So I did install those but in made no difference. Any help would be appreciated.]

aprisma

Hello. Amazing Docs und Workflow. Hope we get 3.0 in future. I just want to tell again. I got crazy with that workflow and reinstalled many times and get always the same red nodes. But if you clear the workflow and load it is fine :-) All working now.

Yukon Cornelius

You need to download those files if you are going to use them. Those names are just on the JSON and were presumably already installed on Aitrepreneurs system. When he saved the JSON for the workflow those are just what he happened to have loaded in there. I would suggest downloading t5xxl_fp16.safetensors t5xxl_fp8_e4m3fn.safetensors ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors ae.safetensors longclip-L.pt And put all of those into the Clip model folder. and actually I just copied most over to the VAE folder too since it got confusing exactly where they need to gosince it seems to change from one node to the next. Then i would go ahead and download a few flux models you like, and a few SDXL models you like and throwing those into the UNET model folder if its flux or in the checkpoints folder if its SDXL basically when you get an error like you are getting it means the file isn't there and is not going to be automatically downloaded. It doesn't actually even mean you need that EXACT file either, for example there are different versions that people have tweaked of each of those files that seem to work equally well if not perhaps better for certain applications. but to be safe you can of course just download what Aitrepeneur had in there when he saved the JSON. The real error you are getting is that whatever you have selected for those slots isn't actually there. You just have to put something that works there and then refresh comfy after they are in their respective folders and you should be able to select them. but don't worry, after you do all that there will be some fresh hell that you need to iron out, lol, the hunt is endless.

Jack Reese

I have the same problem, can you walk me through how you fixed it?

Blue

I'm getting this on startup. Is this normal? Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh"

Jacob Juul

I just can't get it to work. But the worst thing is, that the installation process has affected my StabilityMatrix installation, so it doesn't work - at all... So goodbye

VXLM

Anyone with a solution to this? Id really like to try out pulid. :(

Taiga

I wish you'd get off of this 'one click' thing; it doesn't work. Almost always some download stalls and breaks what was once working. Now, I'm trying to figure out what was supposed to be downloaded and troubleshoot. First it's 'LLM folder doesn't exist', then some download stalls. Now I get these notes: FETCH DATA from: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE] Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh". Your current root directory is: D:\Ultimate FLUX2\ComfyUI_windows_portable\ComfyUI [] []

Armoton69

Same problem. Fixed the naistyles code according to some guide on github, there was a path error. but uninstalling and reinstalling missing nodes brought the problem back again. Would love to get a solution

Mitch Georgieff

I concur with the one click thing. Total failure with a couple different model setups. I do like the concepts and the all in one ComfyUI thing although ComfyUI is somewhat of a pain. Ultimately for all these models I worked with Claude.ai to get everything setup and the failures worked out. I have the ultimate all in one workflow on linux(no gpu overhead like windows) and also the ultimate tts setup in linux.

Aitrepreneur

everything is saved inside the generated images, just drag and drop the image inside comfyui to automatically load the worflow and prompts used to create it

Aitrepreneur

yes normal error no worries, this will not impact anything

Aitrepreneur

This...is a portable install though..so it won't affect anything else

Aitrepreneur

this error is ok, it won't inpact anything and you can just leave it like that

SlyNine

All I can say is I'm sure I followed it right, but even after refreshing the webpage, reloading the workflow, restarting the server, non of that would work until I hit clear in the comfyUI GUI. I suspect a web browser cache issue.

Aitrepreneur

ok, I'll try it out and if it's good, I'll add it to the V3

Aitrepreneur

Send me a DM with more info and explanation of the issue

SlyNine

Oh, no worries now. It's working great now. I was just saying I had to hit the clear button in the GUI and then load the workflow, or else it kept showing missing nodes when I loaded the workflow. Just thought I would mention it incase it helped anyone else, or maybe something to try. Thank you for the workflow. It's awesome.

Mitch Georgieff

Thanks for that. I need to spend some time with it. Several of the nodes fail in the all in one workflow. Many work as expected so I have been able to do a few things.

doodo412

Thanks Janne Kallio, this worked for me, seems I have to manually install it with that command

Jacob Juul

I just want to say, that the fault was mine and mine alone. Without knowing why after intalling the V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL, Stability Matrix made the white screen. F5 is a really good button. I know that now... And by the way: It's a really good installation video you made. Thank you - thank you - thank you.

Aitrepreneur

that's a good point, I might add an option to dl flux fp8 models separately, I left it there because I and some people use both of these models

Bruno Pires

Hi, i trying a lot to install everything, but 2 missing nodes always are there, I already made it from scrath one time, and use the INSIGHTFACE_AUTO_INSTALL.bat and the CUDA_PATH_ONNXRUNTIME_ERROR_FIXER, but everytime i hit this error: Collecting onnxruntime==1.19.2 Using cached onnxruntime-1.19.2-cp312-cp312-win_amd64.whl.metadata (4.7 kB) ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu==1.15.1 (from versions: 1.17.0, 1.17.1, 1.18.0, 1.18.1, 1.19.0, 1.19.2, 1.20.0) ERROR: No matching distribution found for onnxruntime-gpu==1.15.1 D:\AI\NOVA PASTA AI\comfy\ComfyUI_windows_portable\python_embeded>python.exe -m pip install onnxruntime==1.19.2 onnxruntime-gpu==1.15.1 WARNING: Ignoring invalid distribution ~ip (D:\AI\NOVA PASTA AI\comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages) WARNING: Ignoring invalid distribution ~nnxruntime (D:\AI\NOVA PASTA AI\comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages) Collecting onnxruntime==1.19.2 Using cached onnxruntime-1.19.2-cp312-cp312-win_amd64.whl.metadata (4.7 kB) ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu==1.15.1 (from versions: 1.17.0, 1.17.1, 1.18.0, 1.18.1, 1.19.0, 1.19.2, 1.20.0) ERROR: No matching distribution found for onnxruntime-gpu==1.15.1 Looks like he can't find the 1.15.1, idk if it is a CUDA driver problem, my Python is 3.2 btw Any help?

Otaku_Hedgehog

Can you remind me what's the difference between FP8 and GGUF. I have an RTX 3090, which side of the workflow would be better for me to work with? Also, For the Image Generation group, there doesn't seem to be a node for negative prompts. How can we tell ComfyUI to 'not' do something? For example, due to the images used in the Lora training for my OC character, I cannot get her to stop looking at the viewer when I want her to look at the ground for an image that I'm trying to generate. Thanks again for making this AMAZING workflow! ^_^

Randy Monteith

My Reactor node is called "Reactor Fast Face Swap" and your Reactor node is called "ReactorFaceSwap" and your node is twice as long? Are there two different Reactor Nodes or did you rename yours and merge it with some other things?

Aitrepreneur

python 3.2?? you sure about that?? also have you used my 1 click installer from scratch for installing comfyui?

Aitrepreneur

just different types of optimization. gguf Q8 gives you a slight boost in image quality compared to Fp8 (almost no discernable difference tbh) but is twice as slow if you got 24gb of vram, I use FP8 personally, it's much faster. Negative prompts don't really work with flux since it's a distilled model, there is a way to "hack it in" but it will increase the generation time and ressources and it doesn't even work that well, the only way is to just describe more precisely what you want to see in the prompt

Aitrepreneur

yes I merged multiple nodes together and renamed it, so that everything fits into 1 single node

aprisma

For the run pod: do you maybe have an idea why after install on a network drive i cannot install any node via comfyui manager successfully.

Aitrepreneur

check the troubleshooting video post, I show everything from start to finish for runpod as well

aprisma

I had to install the missing nodes again and again. It become less and less error. Maybe bad connection to network drive because I'm often got no such file ....... After 5! Hours it worked

Aitrepreneur

steal? they might steal it only when you stop the pod not when you use it. I personally never stop and leave a pod, I just delete it and redo the whole install from scratch whenever I need it, cheaper this way. I'll soon modify the local and runpod installer to make them more practical so that you don't have to dl everything you might not need

Otaku_Hedgehog

I see, thank you for the explanation.

Rodokmontau

For faceswap i get (IMPORT FAILED) ComfyUI-PuLID-Flux-Enhanced and (IMPORT FAILED) ReActor Node for ComfyUI

aprisma

Just want to add something. The file transfer in your pod (runpod) do not do directory structure. So if someone like me want to move a lot from his lokal pc to the network pod this can help you: curl -fsSL https://raw.githubusercontent.com/filebrowser/get/master/get.sh | bash filebrowser -r /workspace/ -p 8081 -a 0.0.0.0. And then set the 8081 in the list of the pod (edit pod). The pod will reset then but due to network drive all is fine. Login with admin/admin when you connect to 8081 then. PS: Your workflow and POD is perfect. Thank you so much ......

Arthur van Andel

Hi Aitrepeneur, i get the following PULID error: forward_orig() takes from 7 to 9 positional arguments but 10 were given

Vegetables Cut

Hello, I tried the PULID function, but it seems to work on CPU instead of GPU, and it shows this error, is there any solution? error: ApplyPulidFlux

Vegetables Cut

My problem was fixed... but I don't know why. I had another problem which is that my computer only runs on CPU instead of GPU, but after I restart for a few times, it started to work on GPU, and it just went smoothly.

TP

Hi! I get the following error when trying to run the image to video: "Failed to validate prompt for output 1187: * CogVideoDecode 1212:3: - Exception when validating inner node: tuple index out of range". Any help would be highly appreciated!

Aitrepreneur

the cogvideo nodes were updated and it broke the workflow, I just updated the new workflow and uploaded it on patreon, dl it and load it, make sure that you have updated the ComfyUI CogVideoX Wrapper node to the latest version

Aitrepreneur

check out my troubleshooting video, I explain why, at around minute 15 I think

Aitrepreneur

try using the INSIGHTFACE_AUTO_INSTALL.bat CUDA_PATH_ONNXRUNTIME_ERROR_FIXER.bat as explained in the troubleshooting video

Nathan Lorenzana Art

Hello, I would like to know if there is a way to install it on MimicPc?

Randy Monteith

Can you tell me what those nodes are that you Merged?

Jason Hancox

OK, I'm sorry that I am super new to this shit. It's highly frustrating. I have a Pod deployed and running... and the rest is a mess. I uninstalled my ComfyUI so I could start fresh as I could never get it to work anyway. followed all the steps... I got to install missing custom nodes and it just sits there stating it's "Loading missing nodes (cache)" So after about 10 mins I figured ok I must have everything it needs. Moving on to the next step I load the V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW_N20 and get a serious list of missing node types: Missing Node Types When loading the graph, the following node types were not found DownloadAndLoadFlorence2Model (In group node 'workflow>Florence') workflow>Florence Florence2Run (In group node 'workflow>Florence') Text Find and Replace (In group node 'workflow>Florence') LoadFluxControlNet (In group node 'workflow>Flux Controlnet') workflow>Flux Controlnet ApplyFluxControlNet (In group node 'workflow>Flux Controlnet') XlabsSampler (In group node 'workflow>XLabs Sampler') workflow>XLabs Sampler ImpactGaussianBlurMask (In group node 'workflow>Nodes 3') workflow>Nodes 3 ReActorFaceBoost (In group node 'workflow>REACTORFACESWAP') workflow>REACTORFACESWAP ReActorFaceSwap (In group node 'workflow>REACTORFACESWAP') ReActorMaskHelper (In group node 'workflow>REACTORFACESWAP') SAMLoader (In group node 'workflow>Adetailer') workflow>Adetailer UltralyticsDetectorProvider (In group node 'workflow>Adetailer') FaceDetailer (In group node 'workflow>Adetailer') workflow>outpaintflor Int-🔬 (In group node 'workflow>outpaintsize1') workflow>outpaintsize1 JWFloatDiv (In group node 'workflow>outpaintsize1') JWFloatMul (In group node 'workflow>outpaintsize1') Image Blank (In group node 'workflow>outpaintsize1') JWFloatSub (In group node 'workflow>outpaintsize1') JWFloatAdd (In group node 'workflow>outpaintsize1') LayerUtility: ImageBlendAdvance V2 (In group node 'workflow>outpaintsize1') GrowMaskWithBlur (In group node 'workflow>outpaintfluxend') workflow>outpaintfluxend easy imageSize (In group node 'workflow>outpaintfluxend') workflow>faceswap2 workflow>faceswap3 workflow>pullflo LongCLIPTextEncodeFlux (In group node 'workflow>pulidtxt') workflow>pulidtxt Text Concatenate (JPS) (In group node 'workflow>pulidtxt') PulidFluxInsightFaceLoader (In group node 'workflow>pluidlux') workflow>pluidlux PulidFluxEvaClipLoader (In group node 'workflow>pluidlux') PulidFluxModelLoader (In group node 'workflow>pluidlux') ApplyPulidFlux (In group node 'workflow>pluidlux') DF_Get_image_size (In group node 'workflow>vidsupscal') workflow>vidsupscal JWIntegerMul (In group node 'workflow>vidsupscal') JWImageMix (In group node 'workflow>vidsupscal') FILM VFI (In group node 'workflow>vidsupscal') CogVideoImageEncode (In group node 'workflow>videoencod') workflow>videoencod CogVideoSampler (In group node 'workflow>videoencod') CogVideoDecode (In group node 'workflow>videoencod') ShowText|pysssss MiDaS-DepthMapPreprocessor CannyEdgePreprocessor HEDPreprocessor Label (rgthree) Power Lora Loader (rgthree) LoadAndResizeImage DualCLIPLoaderGGUF UnetLoaderGGUF UltimateSDUpscale Text Input [Dream] mxSlider2D Image Comparer (rgthree) mxSlider Fast Groups Bypasser (rgthree) CR SDXL Aspect Ratio DF_Text_Box DownloadAndLoadCogVideoModel VHS_VideoCombine Crop Image TargetSize (JPS) CogVideoTextEncode So THAT is just way outside my ADHD brain's ability to cope with... Where did I go wrong on the ONE click install... it must be something I did I assume.

TP

Thanks, that worked! Great support!

Aitrepreneur

Highly recommend not using MimicPc, I got a heads up saying that the company is not trustworthy, use them at your own risk

Aitrepreneur

have you watched my troubleshooting video post? I show everything step by step

Wonder

Hi there! Can you explain how to get this up and running for a mac user please? Navigating the plaintext doc of your .bat files is too confusing. I already have comfyui up and running, I just need access to your custom nodes and the .json workflow, where can I find this in the main download you provided in 'ComfyUI_windows_portable' ? also does this workflow only work with GGUF models? I want to use the main Flux Dev model

Aitrepreneur

you can try this models installer for mac, I haven't tried it and I don't have any apple products to try myself: https://we.tl/t-YqrrhF9CFH Download that .sh file, put it inside the comfyui folder (the folder where you can see the models folder,etc) open the cmd folder into that folder, use the following commands: chmod +x V2-MODELS-ULTIMATE_FLUX_INSTALL-MAC.sh ./V2-MODELS-ULTIMATE_FLUX_INSTALL-MAC.sh hopefully this works, if not, welp not much I can do since I don't do support for mac and also gguf are just optimized models of the Flux Dev model, there is no reason to use the original model, it uses more vram and there is no difference in quality, but you can also use the fp8 version of the model as well

Wonder

alright well thank you for the help. I didn't manage to get it working, I just was really looking for the workflow.json. I didn't realize the ggufs had no difference in quality, thats great to know thank you!

Aitrepreneur

You can follow the manual install and download all the models and files manually then use the workflow as well

Hank

Hello, I tried the installer multiple times on my machine and on runpod, but in both cases I get to the error "[ERROR] An error occurred while retrieving information for the 'Florence2ModelLoader' node." do I need to download the model myself? how?? thanks

Aitrepreneur

can you send some screenshots? some error logs? dm me the info

Dan Jones

How can I get to negative prompts?

Aitrepreneur

Flux is a distilled model so negative prompts don't work, the only way is to precisely describe what you want to see in the image

aprisma

A question a little out of content. Is it possible somehow add a web browser to your runpod pod? Reason the download from my pc to runpod is very slow. But to github, civitai etc directly much much faster.

talon nightshade

Hi, new here! I'm wondering which version of cuda I should have installed to be able to use this : 11.8, 12.1, 12.4 or something else?

talon nightshade

there is a fix, copy an edited version of naistyler_nodes.py to ComfyUI-Universal-Styler directory in custom nodes.

talon nightshade

This is in my notes from a redit post. this fix works. Make a backup copy of custom_nodes/ComfyUI-Universal-Styler/naistyler_nodes.py ---> naistyler_nodes.py.BAK edit naistyler_nodes.py (in custom_nodes/ComfyUI-Universal-Styler) change every instance of "ComfyUI-NAI-styler" to "ComfyUI-Universal-Styler" (for ubuntu (24.04)) comment out the lines: cls.naistyles_csv = cls.load_naistyles_csv(os.path.join(folder_paths.base_path, "custom_nodes\\ComfyUI-Universal-Styler\\CSV\\naifilters.csv")) cls.naifilters_csv = cls.load_naifilters_csv(os.path.join(folder_paths.base_path, "custom_nodes\\ComfyUI-Universal-Styler\\CSV\\naistyles.csv")) cls.naitypes_csv = cls.load_naitypes_csv(os.path.join(folder_paths.base_path, "custom_nodes\\ComfyUI-Universal-Styler\\CSV\\naitypes.csv")) And add the lines: cls.naistyles_csv = cls.load_naistyles_csv(os.path.join(folder_paths.base_path, "custom_nodes//ComfyUI-Universal-Styler//CSV//naifilters.csv")) cls.naifilters_csv = cls.load_naifilters_csv(os.path.join(folder_paths.base_path, "custom_nodes//ComfyUI-Universal-Styler//CSV//naistyles.csv")) cls.naitypes_csv = cls.load_naitypes_csv(os.path.join(folder_paths.base_path, "custom_nodes//ComfyUI-Universal-Styler//CSV//naitypes.csv")) Note: remember the indention has to be the same as it was before modifications.

Aitrepreneur

no but if you need to quickly upload a large file I use huggingface as a cloud service, upload your file there then use the curl command to quickly download that file onto runpod. the command looks something like this for example a .safetensors file download: curl -L https://huggingface.co/yourusername/thataimodelname/resolve/main/nameofyourfile.safetensors?download=true -o nameofyourfile.safetensors you can use this to very quicly upload a large file from your pc into runpod

Aitrepreneur

well my installer technically uses the comfyui portable version so it doesn't matter, also you can have multiple cuda version installed onto your pc without any issues (for example I got 11.7, 11.8 and 12.1) and everything will work fine

Aitrepreneur

Yes the fix is simple, i just don't do it because I'm lazy since it doesn't inpact the usage😄

talon nightshade

ah looks like 12.1 from the batch file, got it

aprisma

Thank you so much for this of topic thing :-) Btw. I just found there are 2 custom_nodes that download from huggingface, civitai

Nathan Lorenzana Art

Ok, I followed your advice. I installed Runpod, and I got this error: [INFO] Downloading diffusion_pytorch_model-00001-of-00003.safetensors... % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 1186 100 1186 0 0 6690 0 --:--:-- --:--:-- --:--:-- 6700 10 4761M 10 512M 0 0 39.0M 0 0:02:01 0:00:13 0:01:48 40.3M curl: (23) Failure writing output to destination root@18b36d009fe0:/workspace/ComfyUI#

Vegetables Cut

Sorry I faced a new problem... I tried the image to video function, but it doesn't work, below is the CMD message, and I'm using a 4060 laptop GPU: got prompt Failed to validate prompt for output 1191: * CogVideoDecode 1212:3: - Exception when validating inner node: tuple index out of range Output will be ignored Failed to validate prompt for output 1187: Output will be ignored Prompt executed in 0.01 seconds

Zind

what do or where do we place this V2-MODELS-ULTIMATE_FLUX_INSTALL.bat?

Aitrepreneur

read the post man, it's supposed to go in the comfyui folder

George Strnad

Hey guys, my programmer got it installed on our Office 4090 (24gb vram), I'm accessing it through NGROK from my home, but some of the workflows don't work. like FP8 Image gen doesn't work at all but I can get the GGUF image gen working. Secondly, has anyone had success running that PULID? maybe I'm not doing it right, but it just keeps running, for 20 mins, it says 245% completed, so I just stopped it

Nathan Lorenzana Art

I followed your advice (now using MimicPC) I made an account on Runpod, and I got this error (at step#7): what should I do ? I'm new at this platform, I have followed the steps 3 times from start to finish and I get the same error. [INFO] Downloading diffusion_pytorch_model-00001-of-00003.safetensors... % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 1186 100 1186 0 0 6690 0 --:--:-- --:--:-- --:--:-- 6700 10 4761M 10 512M 0 0 39.0M 0 0:02:01 0:00:13 0:01:48 40.3M curl: (23) Failure writing output to destination root@18b36d009fe0:/workspace/ComfyUI#

Pratik Rasam

Is the V2-ULTIMATE_FLUX_ALL-IN-ONE-WORKFLOW.json file attached somewhere else?

Aitrepreneur

what error are you getting for the FP8? Need the error message for all or else I can't troubleshoot it

Paul Tilley

First off, thank you for creating and sharing this fantastic workflow. I'm currently working on setting it up and exploring its features, but I’ve run into an issue when trying to install and use LoRAs. Here’s what I’ve done so far: Downloaded and placed the LoRAs in the LoraModels folder within the custom_nodes/rgthree-comfy directory. Installed all required dependencies via requirements.txt and verified the Python environment is properly configured. Followed the workflow setup instructions, but when I try to use the "Lora Loader" node in the interface, no LoRAs appear in the dropdown or popup menu. Here’s some context that might help: I’ve confirmed the LoRAs are in .safetensors format and verified their integrity. The logs show no errors specific to LoRAs, but I do see a few warnings about missing CSV files and custom nodes failing to load (e.g., "NAI-styles.csv not found"). I’m using the latest version of your workflow and ComfyUI with all dependencies installed. Could you kindly guide me on the following? Is there a specific folder where LoRAs should be placed for your workflow to detect them? Are there additional steps, like setting configurations or running a specific script, to enable LoRAs? If possible, could you share an example or tutorial on integrating LoRAs into this workflow? Thank you in advance for your time and help. I appreciate all the effort that goes into creating and maintaining such a complex tool. Please let me know if there’s additional information I can provide to help debug this.

Aitrepreneur

that's not where you put the loras, you put them in the models/lora folder then just simply select them in the lora loader node

Zind

@Aitrepreneur do the workflows contain a way to caption or train loras for these new models?

Aitrepreneur

no this one, maybe a later one might, altough the text to prompt could technically write the caption for the image I suppose

Micahel DeLeo

I run the video workflow and the video looks really compressed. Any idea why that's the case?

Aitrepreneur

Are you using the latest n20 workflow? Make sure that the crf value is less than 20

Micahel DeLeo

Ah it's actually the upscaler. Should that CRF be less than 20 as well?

Adam

I'm running Q5 gguf because I only have a 3070 (12gb) and it seems to generate deformed hands and body proportions about 70% of the time. Is there somewhere I need to put in negative prompts or am I missing something ???

VirtualVices

I'm getting the : dpt_hybrid-midas-501f0c75.pt.501f0c75b3bca7daec6b3682c5054c09b366765aef6fa3a09d03a5cb4b230853.incomplete I have it downloaded but I dont think it's in the right place :)

Aitrepreneur

not sure, 70% seems a bit much, depends on your prompt I suppose, what prompt did you use?

Aitrepreneur

the dpt_hybrid-midas-501f0c75.pt file should be inside the ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts\lllyasviel\Annotators folder, it should be downloaded automatically

VladoPortos

Hi, where does it install stuff ? and is there an uninstall ?

Aitrepreneur

it installs in the same folder where you put the installer, the folder is called ComfyUI_windows_portable to uninstall it, just delete the folder

VladoPortos

Strange, I do not see the folder, I dropped the .bat file into G:\ run it, and so far I can see my C:\ filling up :) Did a search for the folder and its in: "C:\Windows\System32\ComfyUI_windows_portable" :D

Aitrepreneur

Yes that's probably because you ran the installer as admin, don't do that, run it normally

VladoPortos

Correct, my fault :) I will let it finish and than move the folder... no harm done.

Floris

Hi I'm getting this error when using the IMG to Vid: "CogVideoTextEncode Allocation on device " how can I fix this, I have a 16GB VRAM & RTX4080

Floris

thx, I've set the VRAM to LOW when starting comfyui and it works

Joseph Christensen

I have gotten everything to work now except the image to video upscale. i get the low-res video but it never moves on to #1227 node.

MonoChromancer

I am getting an issue with outpainting. It does the preview but then in comfyui it silently fails and doesn't continue. I followed all the steps troubleshooted all the other issues I encountered, installed it properly (as far as i can tell), the image is in there and fine in the preview, all the connections seem fine. I am guessing/assuming that it is the missing sd_xl_base_1.0_0.9vae.safetensors' that it claims, but ill be trying to see if manually downloading that fixes all of the follow up issues? Will post if that fixes it. In the command prompt I see this: Offloading model... Prompt executed in 8.50 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] * ControlNetApplyAdvanced 981:3: - Required input is missing: image * InvertMask 941: - Required input is missing: mask * VAEEncodeForInpaint 981:4: - Required input is missing: pixels Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored invalid prompt: {'type': 'prompt_outputs_failed_validation', 'message': 'Prompt outputs failed validation', 'details': '', 'extra_info': {}} got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' # 😺dzNodes: LayerStyle -> ImageBlendAdvanceV2 Processed 1 image(s). Prompt executed in 0.97 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' # 😺dzNodes: LayerStyle -> ImageBlendAdvanceV2 Processed 1 image(s). Prompt executed in 1.03 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' # 😺dzNodes: LayerStyle -> ImageBlendAdvanceV2 Processed 1 image(s). Prompt executed in 0.90 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' # 😺dzNodes: LayerStyle -> ImageBlendAdvanceV2 Processed 1 image(s). Prompt executed in 0.86 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' # 😺dzNodes: LayerStyle -> ImageBlendAdvanceV2 Processed 1 image(s). Prompt executed in 0.74 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' WARNING: DreamInputText.IS_CHANGED() got an unexpected keyword argument 'value' Prompt executed in 0.11 seconds got prompt Failed to validate prompt for output 965: * CheckpointLoaderSimple 981:1: - Value not in list: ckpt_name: 'sd_xl_base_1.0_0.9vae.safetensors' not in [] * ControlNetLoader 981:0: - Value not in list: control_net_name: 'diffusion_pytorch_model_promax.safetensors' not in [] * ControlNetApplyAdvanced 981:3: - Required input is missing: image * InvertMask 941: - Required input is missing: mask * VAEEncodeForInpaint 981:4: - Required input is missing: pixels Output will be ignored Failed to validate prompt for output 982:7: Output will be ignored Failed to validate prompt for output 953: Output will be ignored Failed to validate prompt for output 959: Output will be ignored invalid prompt: {'type': 'prompt_outputs_failed_validation', 'message': 'Prompt outputs failed validation', 'details': '', 'extra_info': {}}

Eyes_Wide_Shut

Thank you so much for this all inclusive workflow for Flux! It has been amazing, and I have been using it everyday since you released it! However, can you possibly make a workflow like this for SDXL? I like Flux, but it's difficult to train at the moment, and SDXL already has a large amount of resources ready to go, just need a streamline way to take advantage of them.

Aitrepreneur

what's the error? usually it's surely a vram issue, try using the offload option

Aitrepreneur

have you used the installer from scratch or the model downloader bat file? This should have dl the model automatically

Aitrepreneur

yup I know that this workflow has been greatly appreciated and I'll probably do the same for SDXL based models as well, when? Not sure tbh, there are a lot of future videos already programmed but it's on my list yes

kimitri

I got the same error, did you fix it?

Petar Penchev

Great Stuff ! Works like a charm. One question though : Is it possible to generate video from image longer than 6 seconds even if it is much slower? I tried with offloading but all I got was an error.

Nathan Lorenzana Art

I fixed this error, but then I get another error. I have tried almost 18 times and the last error keeps coming out. The suggestion he gave me was to contact Runpod so that they could support me with that error.

Aitrepreneur

The issue is that I can't replicate the errors on my side either so troubleshooting is basically impossible since it seems like it's an issue on runpod's side, so not much I can do

Aitrepreneur

no you can't because that's how that model was created, the 1.5 (that will be in the V3) can do it though

Timothy Danforth

Hey, first of all, awesome stuff! As to my issue, my network connection got interrupted during install, and now every time I try to re install I get thrown SSL errors for updating python or cloning anything from github.

Aitrepreneur

you can also just open a cmd window then drag and drop the installer inside that window then press enter, this will avoid the window from closing

Aitrepreneur

what are you even doing? the bat file installs EVERYTHING. You don't need to edit or change anything or even do anything...Are you sure that you are launching the bat file correctly and not as admin? What's your PC specs and system?

Timothy Danforth

WARNING: Retrying (Retry(total=4, connect=None, redirect=None, status=None)) after connection broken by 'SSLError(SSLCertVerificationError)1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1006)'))': /simple/facexlib

Timothy Danforth

Never mind. I am an idiot. My service is still too unstable to maintain connection. Pro Tip: The ideal time for large scale downloads is probably NOT going to be during the peak of a winter storm...

MarthSR

You may want to add a note: For those scripts, DO NOT FORCE THEM TO RUN AS ADMIN! Windows gets easily confused as to where is what while doing this, and it'll cause the installs to fail. Just run them from a normal command prompt. :)

Wonder

faceswap and pulid both have 1 node that cannot be found despite everything else working, are these just out of date at this point?

Wonder

"Error. No naistyles.csv found. Put your naistyles.csv in the custom_nodes/ComfyUI_NAI-mod/CSV directory of ComfyUI. Then press "Refresh" Where can I get the naistyles.csv?

Aitrepreneur

well I mean in my videos I show how to use them, just run the file normally. If a file needs to be run as admin then I put it in the name of the file so everyone knows but by default the file run normally when you double click it.

Aitrepreneur

that error is normal, it won't inpact anything

BotBoy

Im getting the same issue..

BotBoy

says (IMPORT FAILED) for ReActor Node and PuLID.

MarthSR

Heheh, was just putting that out there, as techie-inclined people will tend to elevate command prompts to install everything. In this case, that's a no-no. Source: I did that, and went 'surprised pikachu face' when nothing worked. 🤣

Trooperware

Is there a way to improve video quality on Image to video? thanks!

The Photographic Eye

Running Comfyui locally - ran the autoinstall.bat Still getting missing node errors: easy imageSize (In group node 'workflow>outpaintfluxend') workflow>outpaintfluxend CogVideoImageEncode (In group node 'workflow>videoencod') workflow>videoencod CogVideoSampler (In group node 'workflow>videoencod') CogVideoDecode (In group node 'workflow>videoencod') workflowoutpaintsize1 workflowoutpaintfluxend workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpulidtxt DownloadAndLoadCogVideoModel workflowvideoencod workflowvidsupscal CogVideoTextEncode In manager it shows Comfyui CogVideoX Wrapper and ComfyUI Easy Use as 'Import Failed' I've tried repeatedly to reinstall them, but no joy. Here's the log: Total VRAM 12287 MB, total RAM 32698 MB pytorch version: 2.5.1+cu118 Set vram state to: NORMAL_VRAM Device: cuda:0 NVIDIA GeForce RTX 3060 : cudaMallocAsync Using pytorch cross attention [Prompt Server] web root: C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\web Adding C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes to sys.path Could not find efficiency nodes [comfyui_controlnet_aux] | INFO -> Using ckpts path: C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts [comfyui_controlnet_aux] | INFO -> Using symlinks: False [comfyui_controlnet_aux] | INFO -> Using ort providers: ['CUDAExecutionProvider', 'DirectMLExecutionProvider', 'OpenVINOExecutionProvider', 'ROCMExecutionProvider', 'CPUExecutionProvider', 'CoreMLExecutionProvider'] Loaded ControlNetPreprocessors nodes from C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux Could not find AdvancedControlNet nodes Could not find AnimateDiff nodes Loaded IPAdapter nodes from C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus Loaded VideoHelperSuite from C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-VideoHelperSuite ### Loading: ComfyUI-Impact-Pack (V7.12) ### Loading: ComfyUI-Impact-Pack (Subpack: V0.8) ### Loading: ComfyUI-Impact-Pack (V7.12) [Impact Pack] Wildcards loading done. Loaded ImpactPack nodes from C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack [Impact Pack] Wildcards loading done. Traceback (most recent call last): File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 853, in _get_module return importlib.import_module("." + module_name, self.__name__) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "importlib\__init__.py", line 126, in import_module File "", line 1204, in _gcd_import File "", line 1176, in _find_and_load File "", line 1147, in _find_and_load_unlocked File "", line 690, in _load_unlocked File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\loaders\unet.py", line 46, in from .lora_pipeline import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE, TEXT_ENCODER_NAME, UNET_NAME File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\loaders\lora_pipeline.py", line 36, in from .lora_base import LoraBaseMixin File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\loaders\lora_base.py", line 47, in from peft.tuners.tuners_utils import BaseTunerLayer File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\__init__.py", line 22, in from .auto import ( File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\auto.py", line 31, in from .config import PeftConfig File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\config.py", line 24, in from .utils import CONFIG_NAME, PeftType, TaskType File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\utils\__init__.py", line 22, in from .loftq_utils import replace_lora_weights_loftq File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\utils\loftq_utils.py", line 26, in from huggingface_hub.errors import HFValidationError, LocalEntryNotFoundError ImportError: cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 853, in _get_module return importlib.import_module("." + module_name, self.__name__) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "importlib\__init__.py", line 126, in import_module File "", line 1204, in _gcd_import File "", line 1176, in _find_and_load File "", line 1126, in _find_and_load_unlocked File "", line 241, in _call_with_frames_removed File "", line 1204, in _gcd_import File "", line 1176, in _find_and_load File "", line 1147, in _find_and_load_unlocked File "", line 690, in _load_unlocked File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\autoencoders\__init__.py", line 1, in from .autoencoder_asym_kl import AsymmetricAutoencoderKL File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\autoencoders\autoencoder_asym_kl.py", line 23, in from .vae import DecoderOutput, DiagonalGaussianDistribution, Encoder, MaskConditionDecoder File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\autoencoders\vae.py", line 25, in from ..unets.unet_2d_blocks import ( File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\unets\__init__.py", line 6, in from .unet_2d import UNet2DModel File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\unets\unet_2d.py", line 24, in from .unet_2d_blocks import UNetMidBlock2D, get_down_block, get_up_block File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\unets\unet_2d_blocks.py", line 36, in from ..transformers.dual_transformer_2d import DualTransformer2DModel File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\transformers\__init__.py", line 13, in from .prior_transformer import PriorTransformer File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\transformers\prior_transformer.py", line 9, in from ...loaders import PeftAdapterMixin, UNet2DConditionLoadersMixin File "", line 1229, in _handle_fromlist File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 843, in __getattr__ module = self._get_module(self._class_to_module[name]) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 855, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.loaders.unet because of the following error (look up to see its traceback): cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\nodes.py", line 2035, in load_custom_node module_spec.loader.exec_module(module) File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper\__init__.py", line 2, in from .model_loading import NODE_CLASS_MAPPINGS as MODEL_CLASS, NODE_DISPLAY_NAME_MAPPINGS as MODEL_DISPLAY File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper\model_loading.py", line 40, in from diffusers.models import AutoencoderKLCogVideoX File "", line 1229, in _handle_fromlist File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 843, in __getattr__ module = self._get_module(self._class_to_module[name]) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 855, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.models.autoencoders.autoencoder_kl_cogvideox because of the following error (look up to see its traceback): Failed to import diffusers.loaders.unet because of the following error (look up to see its traceback): cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) Cannot import C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-CogVideoXWrapper module for custom nodes: Failed to import diffusers.models.autoencoders.autoencoder_kl_cogvideox because of the following error (look up to see its traceback): Failed to import diffusers.loaders.unet because of the following error (look up to see its traceback): cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) [Crystools INFO] Crystools version: 1.21.0 [Crystools INFO] CPU: AMD Ryzen 7 3700X 8-Core Processor - Arch: AMD64 - OS: Windows 10 [Crystools INFO] Pynvml (Nvidia) initialized. [Crystools INFO] GPU/s: [Crystools INFO] 0) NVIDIA GeForce RTX 3060 [Crystools INFO] NVIDIA Driver: 566.14 Traceback (most recent call last): File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 853, in _get_module return importlib.import_module("." + module_name, self.__name__) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "importlib\__init__.py", line 126, in import_module File "", line 1204, in _gcd_import File "", line 1176, in _find_and_load File "", line 1147, in _find_and_load_unlocked File "", line 690, in _load_unlocked File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\loaders\unet.py", line 46, in from .lora_pipeline import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE, TEXT_ENCODER_NAME, UNET_NAME File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\loaders\lora_pipeline.py", line 36, in from .lora_base import LoraBaseMixin File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\loaders\lora_base.py", line 47, in from peft.tuners.tuners_utils import BaseTunerLayer File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\__init__.py", line 22, in from .auto import ( File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\auto.py", line 31, in from .config import PeftConfig File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\config.py", line 24, in from .utils import CONFIG_NAME, PeftType, TaskType File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\utils\__init__.py", line 22, in from .loftq_utils import replace_lora_weights_loftq File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\peft\utils\loftq_utils.py", line 26, in from huggingface_hub.errors import HFValidationError, LocalEntryNotFoundError ImportError: cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\nodes.py", line 2035, in load_custom_node module_spec.loader.exec_module(module) File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use\__init__.py", line 21, in imported_module = importlib.import_module(".py.{}".format(module_name), __name__) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "importlib\__init__.py", line 126, in import_module File "", line 1204, in _gcd_import File "", line 1176, in _find_and_load File "", line 1147, in _find_and_load_unlocked File "", line 690, in _load_unlocked File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use\py\easyNodes.py", line 17, in from .layer_diffuse import LayerDiffuse, LayerMethod File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use\py\layer_diffuse\__init__.py", line 12, in from .model import ModelPatcher, TransparentVAEDecoder, calculate_weight_adjust_channel File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use\py\layer_diffuse\model.py", line 23, in from diffusers.models.unets.unet_2d_blocks import UNetMidBlock2D, get_down_block, get_up_block File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\unets\__init__.py", line 7, in from .unet_2d_condition import UNet2DConditionModel File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\models\unets\unet_2d_condition.py", line 22, in from ...loaders import PeftAdapterMixin, UNet2DConditionLoadersMixin File "", line 1229, in _handle_fromlist File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 843, in __getattr__ module = self._get_module(self._class_to_module[name]) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\diffusers\utils\import_utils.py", line 855, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.loaders.unet because of the following error (look up to see its traceback): cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) Cannot import C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use module for custom nodes: Failed to import diffusers.loaders.unet because of the following error (look up to see its traceback): cannot import name 'LocalEntryNotFoundError' from 'huggingface_hub.errors' (C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\errors.py) ComfyUI-FreeMemory nodes loaded: Image, Latent, Model, and CLIP versions available. [ComfyUI-HakuImg]-|21:47:01|-INFO: Check HakuImg requirements [ComfyUI-HakuImg]-|21:47:01|-INFO: Check HakuImg requirements done [ComfyUI-HakuImg]-|21:47:01|-INFO: Check HakuImg submodule module: PixelOE [ComfyUI-HakuImg]-|21:47:01|-INFO: Check HakuImg submodule module done ### Loading: ComfyUI-Impact-Pack (V7.12) [Impact Pack] Wildcards loading done. ### Loading: ComfyUI-Inspire-Pack (V1.9) Total VRAM 12287 MB, total RAM 32698 MB pytorch version: 2.5.1+cu118 Set vram state to: NORMAL_VRAM Device: cuda:0 NVIDIA GeForce RTX 3060 : cudaMallocAsync theUpsiders Logic Nodes: Loaded ### Loading: ComfyUI-Manager (V2.53) ### ComfyUI Revision: 2923 [6d1a3f7d] | Released on '2024-12-15' [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\timm\models\layers\__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning) Please 'pip install xformers' Nvidia APEX normalization not installed, using PyTorch LayerNorm Please 'pip install xformers' Nvidia APEX normalization not installed, using PyTorch LayerNorm [ReActor] - STATUS - Running v0.5.2-a2 in ComfyUI Torch version: 2.5.1+cu118 ------------------------------------------ Comfyroll Studio v1.76 : 175 Nodes Loaded ------------------------------------------ ** For changes, please see patch notes at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/blob/main/Patch_Notes.md ** For help, please see the wiki at https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes/wiki ------------------------------------------ LOADED 6 FONTS No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate' ███╗ ███╗ █████╗ ██████╗██╗ ██╗██╗███╗ ██╗███████╗ ████╗ ████║██╔══██╗██╔════╝██║ ██║██║████╗ ██║██╔════╝ ██╔████╔██║███████║██║ ███████║██║██╔██╗ ██║█████╗ ██║╚██╔╝██║██╔══██║██║ ██╔══██║██║██║╚██╗██║██╔══╝ ██║ ╚═╝ ██║██║ ██║╚██████╗██║ ██║██║██║ ╚████║███████╗ ╚═╝ ╚═╝╚═╝ ╚═╝ ╚═════╝╚═╝ ╚═╝╚═╝╚═╝ ╚═══╝╚══════╝ ██████╗ ███████╗██╗ ██╗ ██╗███████╗██╗ ██████╗ ███╗ ██╗███████╗ ██╔══██╗██╔════╝██║ ██║ ██║██╔════╝██║██╔═══██╗████╗ ██║██╔════╝ ██║ ██║█████╗ ██║ ██║ ██║███████╗██║██║ ██║██╔██╗ ██║███████╗ ██║ ██║██╔══╝ ██║ ██║ ██║╚════██║██║██║ ██║██║╚██╗██║╚════██║ ██████╔╝███████╗███████╗╚██████╔╝███████║██║╚██████╔╝██║ ╚████║███████║ ╚═════╝ ╚══════╝╚══════╝ ╚═════╝ ╚══════╝╚═╝ ╚═════╝ ╚═╝ ╚═══╝╚══════╝ Please 'pip install xformers' Nvidia APEX normalization not installed, using PyTorch LayerNorm [rgthree-comfy] Loaded 42 extraordinary nodes. 🎉 WAS Node Suite: OpenCV Python FFMPEG support is enabled WAS Node Suite Warning: `ffmpeg_bin_path` is not set in `C:\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\was-node-suite-comfyui\was_suite_config.json` config file. Will attempt to use system ffmpeg binaries if available. WAS Node Suite: Finished. Loaded 218 nodes successfully. "Art should disturb the comfortable and comfort the disturbed." - Cesar Cruz I've also had issues installing Xformers - it fails no matter what I do when trying to build wheels. Are the two related? Thanks

Robert Morton

Running comfyui locally (no hardware issues, powerful pc) Getting this error when trying to run the img2video Failed to validate prompt for output 1227: * UpscaleModelLoader 1236:0: - Value not in list: model_name: '4x-UltraSharp.pth' not in ['4x-ClearRealityV1.pth'] Output will be ignored the low res version renders out as expected but no upscaled version, please advise how to fix this please.

Mir

I have the same problem :(

Nic Montemayor

I'm getting an error when trying to use Pulid. SamplerCustomAdvanced forward_orig() got an unexpected keyword argument 'attn_mask'

Pearce

on runpod can't install was node suite or comfyui layer style

Kristof

Failed to validate prompt for output 1227: * CogVideoSampler 1237:1: - Return type mismatch between linked nodes: feta_args, received_type(VAE) mismatch input_type(FETAARGS) * CogVideoDecode 1237:2: - Required input is missing: vae Output will be ignored what does it mean? image to video part.

Ylong Rusk

Hello, please help me with this, as soon as this V2-MODELS-ULTIMATE_FLUX_INSTALL.bat finished installation i cant open comfyui anymore, i get this error: H:\comfy>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build [START] Security scan [DONE] Security scan ## ComfyUI-Manager: installing dependencies done. ** ComfyUI startup time: 2025-01-07 14:58:36.152730 ** Platform: Windows ** Python version: 3.11.9 (tags/v3.11.9:de54cf5, Apr 2 2024, 10:12:12) [MSC v.1938 64 bit (AMD64)] ** Python executable: H:\comfy\python_embeded\python.exe ** ComfyUI Path: H:\comfy\ComfyUI ** Log path: H:\comfy\ComfyUI\user\comfyui.log [START] Security scan [DONE] Security scan object address : 0000016805D9A030 object refcount : 2 object type : 00007FFB6837A8D0 object type name: AttributeError object repr : AttributeError("type object 'datetime.datetime' has no attribute 'datetime'") lost sys.stderr H:\comfy>pause Press any key to continue . . .

Ylong Rusk

I want also to add that python version there is not the same as in my system? I have 3.9.10 installed but that shows a newer version there when i try to open comfyui, you think that could be the problem? and if yes how can I remove it? I come from the video of omnigen, that maybe made a new python version...

Leunis van Rooyen

If I have nothing on my PC and I want to just run the "1 click installer", which file do I run? V2-MODELS-ULTIMATE_FLUX_INSTALL or the one called V2-ULTIMATE_FLUX-COMFYUI-MANAGER_AUTO_INSTALL

Leunis van Rooyen

I did those and when I run the nvidia batch file I get this: [START] Security scan [DONE] Security scan Failed to execute startup-script: D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\prestartup_script.py / module 'folder_paths' has no attribute 'get_user_directory' Prestartup times for custom nodes: 1.0 seconds (PRESTARTUP FAILED): D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager Total VRAM 16376 MB, total RAM 31895 MB pytorch version: 2.3.1+cu121 Set vram state to: NORMAL_VRAM Device: cuda:0 NVIDIA GeForce RTX 4080 : cudaMallocAsync Using pytorch cross attention Traceback (most recent call last): File "D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1931, in load_custom_node module_spec.loader.exec_module(module) File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\__init__.py", line 8, in import manager_server # noqa: F401 ^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\glob\manager_server.py", line 13, in import git ModuleNotFoundError: No module named 'git' Cannot import D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager module for custom nodes: No module named 'git' Traceback (most recent call last): File "D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1931, in load_custom_node module_spec.loader.exec_module(module) File "", line 940, in exec_module File "", line 241, in _call_with_frames_removed File "D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux\__init__.py", line 1, in from .pulidflux import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS File "D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux\pulidflux.py", line 10, in from comfy.ldm.flux.layers import timestep_embedding ModuleNotFoundError: No module named 'comfy.ldm.flux' Cannot import D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux module for custom nodes: No module named 'comfy.ldm.flux' Import times for custom nodes: 0.0 seconds: D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py 0.0 seconds (IMPORT FAILED): D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux 0.0 seconds (IMPORT FAILED): D:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager

Yvette Legrand

it is possible to create a video from a photo with this webUI? or do you intend to add another post about a good local install AI that can do that simply generate video and not just photo

Michael Stone

I updated Comfy, and it broke.... well lots of things. So I'm trying to do a clean install, and ran into 2 problem. I managed to fix one. Running the batch file I got this error at the end: Collecting onnxruntime==1.19.2 Using cached onnxruntime-1.19.2-cp312-cp312-win_amd64.whl.metadata (4.7 kB) ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu==1.15.1 (from versions: 1.17.0, 1.17.1, 1.18.0, 1.18.1, 1.19.0, 1.19.2, 1.20.0, 1.20.1) ERROR: No matching distribution found for onnxruntime-gpu==1.15.1 [ERROR] Failed to install onnxruntime, onnxruntime-gpu, or InsightFace. I changed the 1.15.1 for onnxruntime-gpu to the same 1.19.2 version and it installed fine after that. Now I'm attempting to update all the missing nodes, and it's not doing anything. I get the spinning circle in the screen, and nothing happens. The last lines I get in the command window are: FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\832903789_extras.json [DONE] FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\1742899825_extension-node-map.json [DONE] FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\2233941102_nodes_page_1_limit_1000.json [DONE] nightly_channel: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/cache FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\1514988643_custom-node-list.json [DONE] FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\746607195_github-stats.json [DONE] FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\832903789_extras.json [DONE] FETCH DATA from: M:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\cache\1742899825_extension-node-map.json [DONE] ....and it just stops and does not move.

Nils Johan Anti

anyone got a solution to this?, i've got the same problem

Michael Stone

Update.... I went back to my old install that had stopped working (I had just renamed the directory), updated everything again, and it's working again now. The new install still won't install missing nodes, but whatever was the problem with the old one got fixed in an update.

Daniel A Perez

I am using runpod, and I get an error and cannot install the ReActor Node for ComfyUI. When I go to Github it has been taken down. Anywhere else where this can be downloaded?

Luis

Hi K! Just a wish, it is possible for they next versión to add a module to change facial expressions of existing photo maintaining the face? Thanks

Jahon Rockny

whenever i try the image to video i get this error: Failed to validate prompt for output 1227: * CogVideoSampler 1237:1: - Return type mismatch between linked nodes: feta_args, received_type(VAE) mismatch input_type(FETAARGS) * CogVideoDecode 1237:2: - Required input is missing: vae Output will be ignored

Morpheus

Hey guys Has anyone figured out what is causing the following error when attempting to run Image To Video? Failed to validate prompt for output 1227: * CogVideoSampler 1237:1: - Return type mismatch between linked nodes: feta_args, VAE != FETAARGS * CogVideoDecode 1237:2: - Required input is missing: vae Output will be ignored Failed to validate prompt for output 1218: Output will be ignored Prompt executed in 0.01 seconds Can't figure out how to fix it? does anyone know how to fix this

Papalap

me too -.-

Papalap

found any help? got the same problem

Gamearc

ReActorMaskHelper Weights only load failed. This file can still be loaded, to do so you have two options, ?[1mdo those steps only if you trust the source of the checkpoint?[0m. (1) In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load with `weights_only=True` please check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL ultralytics.nn.tasks.DetectionModel was not an allowed global by default. Please use `torch.serialization.add_safe_globals([DetectionModel])` or the `torch.serialization.safe_globals([DetectionModel])` context manager to allowlist this global if you trust this class/function. Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html. Any fix?

arif hidayat

miss match... moving feta_arg connection from videoencod to vae >>> Feta_arg > Vae in videoencod too

Pizza Yap

Hi if I already have installed comfyui in runpod, do I follow the "IF YOU ALREADY HAVE COMFYUI INSTALLED" or the "IF YOU ARE USING RUNPOD" steps?

Jeremiah Everett

This is all I get when I run the nvidia run bat. C:\ComfyUI_windows_portable>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build Traceback (most recent call last): File "C:\ComfyUI_windows_portable\ComfyUI\main.py", line 80, in import execution File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 11, in import nodes File "C:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 21, in import comfy.diffusers_load File "C:\ComfyUI_windows_portable\ComfyUI\comfy\diffusers_load.py", line 3, in import comfy.sd File "C:\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 5, in from comfy import model_management File "C:\ComfyUI_windows_portable\ComfyUI\comfy\model_management.py", line 119, in total_vram = get_total_memory(get_torch_device()) / (1024 * 1024) ^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable\ComfyUI\comfy\model_management.py", line 88, in get_torch_device return torch.device(torch.cuda.current_device()) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\cuda\__init__.py", line 1026, in current_device _lazy_init() File "C:\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\cuda\__init__.py", line 363, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled C:\ComfyUI_windows_portable>pause Press any key to continue . . .

Daniele Casali

at point 9, after clicking Install Missing Custom Nodes, I only get a black list and a red notice Failed to get custom node list. can you help me?