Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

KONTEXT NEW MULTIMODAL AI IMAGE EDITING KING WORKFLOW! (Patreon)

Published:
2025-07-02 19:39:40
Edited:
2025-07-04 17:11:44
Imported:
2026-06
Tags:

Downloads

Content

Edit 04th July: I reuploaded the workflow by adding another lora loading method in the nunchaku model workflow because some loras don't work with the nunchaku lora loader WHILE OTHERS don't work with the normal lora loader...probably has something to do with the trainer the person used to make the lora. So basically only have ONE lora loader enabled and have the other disabled. If one loader is not working, try the other one and vice versa....

Hey everyone! I've just released a new video on "THE NEW MULTIMODAL AI KING!", called "KONTEXT". Simply THE MOST POWERFUL AND INSANE AI IMAGE EDITING MODEL EVER! A brand new 12-BILLION-PARAMETER model from the makers of Flux called “Kontext”. A multimodal model that is able to automatically edits images with incredible precision. The amount of things you can do with this model is just mind blowing AND by using the brand new Nunchaku model, can generate images IN SECONDS!!!… It's just amazing!

You can check out the video right here: https://youtu.be/Lsz3tlbBaqM

I also created a 1-click installer for the Webui for my Patreon supporters for Local AND Runpod, that you can find right here:1-Click INSTALL KONTEXT AI IMAGE EDITING KING ComfyUI WebUI! | Patreon

Here the link to the manual install guide: https://pastebin.com/jTGgsFbA

Down below you will also find the FLUX_KONTEXT_IMAGE_EDITING.json for free for you to use!

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😉!Ps: Sorry for being very quiet again, some of you kinda know the drill by now....it's nothing new...spend a few days in the hospital and was bedridden for a while so thank you all for being patient! 🫶

Files

Comments

MonkeyDanceParty

Great stuff, hope you are doing better. Really like your videos.

Afro_Fail

I'm not seeing the FLUX_KONTEXT_IMAGE_EDITING.json is it in this post? amazing work by the way, this is been very confusing trying get an efficient workflow through reddit searching.

rubz

Great! I've spend the whole week experimenting with different test and flows. Gonna try this one asap. Hope you are ok

Manic

NunchakuFluxLoraLoader and NunchakuFluxDiTLoader can't be downloaded for me no mater how many times I try.

Aitrepreneur

Have you done the install from scratch or from an existing install?

Manic

exsisting install. should I do it by itself?

Aitrepreneur

it's advised to use the KONTEXT-COMFYUI-MANAGER_AUTO_INSTALL.bat installer since it's easier but the other installer can work as well, what exactly is the error you're getting?

Dattprasad Salunke

sorry, as iIam new, I could not find FLUX_KONTEXT_IMAGE_EDITING.json

Mathilde Leroux

Missing Node Types When loading the graph, the following node types were not found NunchakuFluxDiTLoader NunchakuFluxLoraLoader

Ragscoon Raccoon

Very Lame that you gatekeep this info and tools behind a paywall

Gillian

I had the experimental working fine but I changed something and get arror 'SamplerCustomAdvanced The size of tensor a (768) must match the size of tensor b (128) at non-singleton dimension 2' cannot figure out what needs changing?

Justin

He's the one making them for our convenience. Why shouldn't he make some money for his time?

Bobby Boe

Some information for you: As many influencers that started with Comfy, you might not have noticed that there is a version called ComfyUI Desktop. This version makes it much easier to start with ComfyUI and to install new features. There is no more need to have complicated git pulls etc. It is way easier now. It is only confusing, that many influencers are still teaching complicated ways. In many cases, ComfyUI can even install the model automatically, when starting the workflow for the first time. The draw back is, that ComfyUI Desktop creates some virtual environment automatically, which makes it diffucult to manually install things... Unfortunatly, there arent influencers teaching workarounds for that... so for example, I am still struggling for nunchaku (its not installing). I hope my information was helpful, keep up the good work my friend!

Aitrepreneur

Have you done a full reinstall of comfyui and done the nunchaku wheels install ?

Aitrepreneur

The workflow is free and I even make a document explaining how to install everything. Compare that to some creators that do nothing for the free folks members.

Wiktor Waszak

Hi, i run this on runpod. When trying to install nunchaku nodes i get error message "import failed" and nunchaku is not available. After restart and refresh it again shows a message that nunchaku nodes are missing

Aitrepreneur

I disagree, I'm using literally the easiest version possible for most people since it's the portable version, so there is no need to manage incorrect python versions or anything, everything is already contained in the python embedded folder. My videos and installers are always made with complete beginners in mind, using the desktop version is anything but ideal and is less flexible for most beginners users.

Gillian

yes I changed it probably to a number non divisible by 256? but when I changed back to the default the error remained. I had to restart to get back.

Omar Aloulou

Hi, I’ve been trying to use a workflow that depends on the nodes NunchakuFluxLoraLoader and NunchakuFluxDiTLoader, but ComfyUI shows an error saying these node types are missing. I’ve already reinstalled ComfyUI and made sure all required extensions are up to date. If they are deprecated or renamed? Or if there are additional steps to get them working? Thanks a lot for your help!

Bobby Boe

Hmm, ok, I never used the portable version. But the desktop version came with a "normal" windows-installer, there was no python-checking... The best, if there is a new model (which is supported), you can just load the workflow, and the models are automatically downloaded and placed in the right folders... very comfortable. I did not pip install anything, no cmd prompts at all. Only I have to admit, when things are not running (link nunchaku right now for me) this is less flexible.

Ixalon

I agree with the portable version being easier for new and advanced workflows. I tried using the desktop version for a while, but installing things like Triton and Sageattn were difficult and some of the newer workflows were difficult or impossible to use. Installing the portable with the oneclick installer once then using update scripts with new things that come out is very convenient and easy to update.

SUTY THE REAL

Tried Desktop version, absolutely dislike using it

SUTY THE REAL

i got the error when second image was empty but enabled

SUTY THE REAL

you can always follow the Manual Installation Guide whats linked in hes Vods and stop being lazy >)

Brad

Very Lame you think he should do stuff for free because you ask him

Aitrepreneur

dm me some screenshots of the error and screen

Xpicto

I used the Desktop version for the first time the other day, installed a custom node from the manager... broke everything, blank screen. Standard logging not as visible as usual so was harder to debug etc. Went back to portable version.

Xpicto

Gimme gimme gimme, free free free, now now now. Your efforts mean nothing to me.

The Capn

Link is dead, I found the safetensors file already, but it doesn't do anything in the workflow? any tips of what might be going wrong? (maybe i have the wrong trigger phrasing)

Albion22

where did you find the file ? I've been looking all over for It bit it has been removed from the face of the earth ! Also for the trigger words you can still read it in the huggingface page, even though the links have been removed. Found it. Trigger words are what I said they were.

Aitrepreneur

I'm aware of the issue, I'm trying to find a solution. No worries I got them all saved up locally

firstname lastname

Have you thought about doing a workflow that allows you to do a mask so that you can ensure only what you want changed is changed?

Ech0_Hue

Is all the models and files linked in the manual installation safe?

Mikael Svenson

Do you plan to add an inpaint flow as well? Seen some others do this with Kontext. Awesome work!

GlobalGrooveWORLD

Please don't talk nonsense if you don't understand how awesome this model is in T2I. The old flux model didn't even come close. The quality of the T2I and PRO models is such that they often outperform REVE, IDIOGRAM, and ImageN. At the same time, the DEF model is also not bad and also outperforms the old flux by a factor of several. In Comfy, you just need to find the right samplers, and Forge even works with D...2++. So just walk away and let the author answer.

TomiTom1234

I am in trouble. I get the error message "AssertionError: Torch not compiled with CUDA enabled" I found that my CUDA ver. is 12.9 , but when I applied the command "python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu129" An error in red appears that says: "ERROR: Could not find a version that satisfies the requirement torch (from versions: none) ERROR: No matching distribution found for torch" What should I do?

Aitrepreneur

I tried it, the model IS good, comparable to Flux. Similar style, sometimes better, sometimes worse but pretty much on the same level. However that is compared to the base model though, Flux has a bunch of finetunes and loras that Kontext doesn't have so until we have a true Kontext finetune Flux is still ahead.

Aitrepreneur

I tried it, inpainting with Kontext is either not great or kinda useless compared to the Flux fill model that is MADE for inpainting. The whole point of Kontext is that you don't need to do a mask, if you do a mask, might as well use the Flux fill model (which is an amazing model too btw) I will do an updated video on Flux workflow soon though

Aitrepreneur

cu129? I wasn't even aware this was available. You are not in trouble What's your GPU? have you followed the instruction here: Just go inside the python embedded folder, click on the folder path, type cmd press enter, this will bring the command prompt window and there type: python.exe -m pip uninstall torch python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128 --->replace cu128 by your cuda version (if you don't know your cuda version just open a new cmd window and type nvcc --version it will then tell you at the end Cuda compilation tools, release 12.8, V12.8.93 or something different like 12.6, in that case just use cu126 instead of cu128)

TomiTom1234

Yes I did this instructions. My GPU is 5070 Ti (16 GB). The error appeared when I wanted to perform this last command: python.exe -s -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu129 I changed 1298 to 129. But it seems 12.9 isn't supported as I understood from that command. My ComfyUI doesn't even start anymore because of this issue.

Sean Rowe

I have ComfyUI installed on my system already using StabilityMatrix. I can't get the Nunchuka nodes installed for some reason. It keeps failing to install properly but I don't know why. It says I need "NunchakuFluxDitLoader" and "NunchakuFluxLoraLoader". I've tried installing both through the manager a bunch of times but after every restart it still says I need to install them. Any tips?

Aitrepreneur

Well I'm not sure what StabilityMatrix uses to install comfyui, my installers are using the portable version to work between them. One thing you can do, is to make sure that you got at least the nunchaku custom node installed via the manager and then load this special workflow : https://we.tl/t-kdIzCARCa7 To automatically download and install the nunchaku wheel. Just load this inside comfyui and then click run (as if you are generating an image). If it works, great, you can restart comfyui via the manager and reload the workflow.

TomiTom1234

It says 12.9 Here is what comes out after typing the command asking for the Cuda version: E:\AI\ComfyUI_windows_portable\python_embeded>nvcc --version nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2025 NVIDIA Corporation Built on Wed_Apr__9_19:29:17_Pacific_Daylight_Time_2025 Cuda compilation tools, release 12.9, V12.9.41 Build cuda_12.9.r12.9/compiler.35813241_0

Aitrepreneur

Indeed so the command with cu129 should work then...given that you first uninstall torch like i said in the previous comment, DM me maybe some screens (or even better a short video recording of you doing the whole step) so I can at least confirm that everything is done in order. Worst case scenario you can just reinstall comfyui from scratch with my installer somewhere else then move all models inside the new install

TomiTom1234

Here is a screenshot of the command prompt, when I uninstall the torch, then install/upgrade torch for my cuda version, which is 12.9.. Then at the end I show the cuda version by applying the nvcc --version command. https://ibb.co/C3WZgBrw I hope there is an easy solution rather than installing the ComfyUI from scratch then move everything to it, because the size of the files there are more than 350 gb. ======== EDIT: ======== I solved the issue by installing the torch and using cuda128 instead of the version cuda129, which I found as a version in my machine. Thank you for your great work and time.

TomiTom1234

Nope, it seems only ComfyUI is working, but any other workflow isn't working. I mean I tried different workflows which worked before, but they drive me in many issues. I guess, I messed up the whole ComfyUI.

TomiTom1234

OK I decided to install it from scratch and save myself some headache. Now it is working and smooth. Thank you for your hard work.

Sean Rowe

You are the best! Thanks that fixed the problem for me.

TsT65

Hello, I used the upgrade batch file for my existing ComfyUI Portable. How do I find the nodes "ReferenceLatent", "FluxKontextImageScale" and "ImageStitch"? I don't find them via comfy manager and "install missing nodes" doesn't find them.

Alexander V Morawski

I'm having some difficulty when using 2 images, I usually just get an output of the two images stiched together in the middle or I loose all the details and it ignores my prompt, I know it's working because when I use 1 image it for the most part does what I ask, and the guide doesn't seem to explain how to use 2 images

Aitrepreneur

Depends on the prompt and what you are trying to do, for some thing you need to choose a specific resolution when generating (activate custom resolution)

Kappak Amaru

hello sit-this is working with nvidia rtx 3050 8gb?

Cory Fitzgerald

I just tried the one click installer and ran into a lot of issues I got them all fixed except these two. NunchakuFluxDiTLoader NunchakuFluxLoraLoader I am not sure exactly where I need to go to get these, which ones to actually download, or where to put them

SpudMango

I ran the attached file. It said it updated. When I refreshed I still got the same error. Missing NunchakuFluxDiTLoader and NunchakuFluxLoraLoader. I tried running the wheel installer again, and now its giving me a new error even for the wheel installer saying Nunchakuwheelinstaller is a missing node.

mak

same here, run the workflow provided multiple times no luck no fix

Aitrepreneur

have you used the comfyui installer from scratch? if you have already a previous version you need to use the update bat file inside the update folder to get the latest comfyui update. Then once you got the nunchaku custom nodes installed, you can run the special workflow to install the wheel

Cory Fitzgerald

So this is the only comfyUI installed on the machine I was installing it on. I have another PC that is running the AI generators you put out a video on a few days ago. I can't even get to drop the workflow in. It pops up error as soon as I open up the browser windows saying I need the Nunchuk nodes

SpudMango

I tried from scratch in a completely different comfyui folder. Do I need to completely remove comfyui from my PC somehow? Because I made multiple install folders for different times I have used your installations and updated all of them. For example I have a comfyui folder I used just for KONTEXT-COMFYUI-MANAGER_AUTO_INSTALL.bat and I tried a different comfyui folder with KONTEXT-MODELS-NODES_INSTALL.bat. And now, no matter which installation I choose, the nunchakuwheelinstaller still gives me the same error.

Aitrepreneur

Yup an update just wrecked the nunchaku install, I've updated the installers, redownload them from the page and preferably do an install from scratch

TomiTom1234

I have now issues when loading the json. I get NunchakuFluxDiTLoader NunchakuFluxLoraLoader are missing, and no matter how much I try to download the missing from the manager, no luck. I downloaded https://www.mediafire.com/file/hxdnzksadv0arg5/install_wheel.json file from that link you provided, it did download successfully, but after opening the ComfyUI still the missing loaders. I can use the FP8 but the Nunphy is missing. Any way to fiy the problem? This happened after I updated the ComFyUI.

StudiAI

This solution worked for me. (Nunchaku FLUX DiT Loader e Nunchaku FLUX Lora Loader not working) After reinstalling, everything was working fine — but when I tried to install additional nodes, I got errors with both nunchaku loader nodes. Here’s what I did: I deleted the following folders from custom_nodes: ComfyUI-GGUF ComfyUI-KJNodes ComfyUI-Crystools rgthree-comfy ComfyUI-Manager I reinstalled everything using the KONTEXT-MODELS-NODES_INSTALL.bat file. Since I had already downloaded the large model files before, the process was faster — and all the other nodes I had previously updated remained working properly.

Aitrepreneur

Have you used the updated installers, just updated them 2 days ago

Aitrepreneur

I have updated the installers, so you didn't really have to do all of that. That being said, nunchaku nodes are an absolute shitshow to install and maintain correctly, always having to use the special install wheel workflow to update the wheel each time it breaks so yeah I feel you, got a similar issue on my side a few days ago.....

The Capn

Any updates on a solution for nsfw in the kontext workflow? I have the loras but they don't do anything in the workflow still. any nsfw prompting just leads to a 1:1 unchanged image

Ech0_Hue

Can you update the manual installation guide as well?, because i tyed the install wheel and it i keep getting same error

Kankatzu

no idea what the fuck is up with the nunchaku devs... like... are these the same people that work for EA or why do they deliberately make the life of normal people hard...? Cause this ain't working either and the Qwen Image Editor... well its to stylized, wich in turn makes it useless for photo editing... Kontext was far better as it really only changed what I told it to change and kept the rest of the composition...

james beal

I did. At least if the latest is in your runpod template. Seems to download and install most everything on a new pod, but same issues everyone else seems to experience with Nunchaku. I mean has anyone been able to get this to work on Runpod? After your template installs on RP, and install missing nodes, and then attempting to find 'NunchakuWheeIntaller... it is a recognized in the flow as a missing node, but ComfyUI manager cannot find it. I mean is this seriously a legit model because seriously almost impossible to even install. I feel like people have moved on from this.