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

HIDREAM EASY TEXT-TO-IMAGE WORKFLOW! (Patreon)

Published:
2025-05-05 18:36:50
Edited:
2025-05-06 14:04:15
Imported:
2026-06
Tags:

Downloads

Content

EDIT: TO USE THE WORKFLOW DO NOT FORGET TO UPDATE COMFYUI! DO NOT use the manager to update comfyui, the manager has been broken for months, to update comfyui correctly, go inside the update folder and run the update_comfyui_and_python_dependencies.bat file (do not forget to continue the install once it asks you to press any button) then, once the update is done, relaunch comfyui and re drag and drop the workflow inside.Hey everyone! I've just released a new video on "THE NEW IMAGE GEN KING ! ", called "HIDREAM". Simply THE BEST AND MOST POWERFUL AI TEXT TO IMAGE AI MODEL TO DATE! It's just amazing!

You can check out the video right here: https://youtu.be/OeKRc2CjAmA

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 HIDREAM TEXT-TO-IMAGE NEW AI KING ComfyUI WebUI! | Patreon

Here the link to the manual install guide: https://pastebin.com/hvU8w7y5

Down below you will also find the HIDREAM_EASY-WORKFLOW.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 the lack of news lately, I've already talked about it before, so no need to talk about it again but thanks for your support and patience, much love❤️

Files

Comments

rubz

Really, looking forward to giving this one a spin! Take care !

Sense Departed

hmm, The installer completes, and there is nothing in the workflow when its dropped into comfyui, no errors, just nothing there

Audiogus

Got comfy installed using the one click installer. Seems to load Comfy OK but when I drop the .json into comfy it is completely blank.

El Cabeleira

Could anyone please tell me where I can get the "workflowloader" node from? I can't find it anywhere under modules and on github so I'm probably not searching for the right thing.

ExusOne

You need first to make full update of ComfuUI by going to update folder and choose update_comfyui.bat file. Later try to load it and zoom out as much is possible. On my place the workflow with one missing node from imagetoimage was on down-right side

Art4Ever

I also have this issue. Latest install of KJ nodes and ComfyUI and still I get this error. Please let us know the solution. Thank you :)

Audiogus

Thanks for the tip! But all I get is "stashing current changes nothing to stash creating backup branch: backup_branch_2025-05-05_15_03_34 checking out master branch Traceback (most recent call last): File "H:\HiDream\ComfyUI_windows_portable\update\update.py", line 66, in ref = repo.lookup_reference('refs/remotes/origin/master') ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ KeyError: 'refs/remotes/origin/master' Press any key to continue . . . " This is with the one click installer from the patreon post above.

Robb

I'm having the same issue. Everything downloads but then when I load the workflow it's blank. I'm going to start waiting a day or two after you release these videos because it never works the first time :/

Art4Ever

I found a way around it, after I updated comfyui to latest and KJ Nodes as well, I loaded this workflow, got the message for the "workflowloader" then I opened the previous samelike workflow for Flux for Aitrepreneuer and then dropped again the Hi-Dream workflow and it was working fine for me. No more missing node msgs. I hope that works for you too.

Greg

For those saying it looks blank, try clicking the "Fit View" button on the bottom right.

Jamie1994

Indeed, hope the fix will be comming :)

El Cabeleira

I'll try updating KJ Nodes since I didn't do that, I just tried dropping the workflow in again. Already have the latest comfyui so that shouldn't be the issue. Thank you for the tip :)

Art4Ever

You are welcome. Use the "git pull" from command prompt at the KJ nodes directory, when you are updating comfy-nodes. After restarting everything (after updates), remember Load the previous Flux workflow, and after that try again this workflow. Well I hope it works now. Hi-Dream is really a great model.

Jamie1994

Thats the solution. Thank you!!

theworldofanatnom

literly everything loaded fine except for the inpaint node, its red and just has "workflowhiinpai" and all the data inputs are "unknown" and idk why. comfy auto updates and and i updated all my nodes and restarted but its still not loading

Midnight

Unexpected architecture type in GGUF file, expected one of flux, sd1, sdxl, t5encoder but got 'hidream' I am getting this error. I used the model and node downloader. I choose option 2 for 12-16GB I think it said.

Tomasz Czekaj

on right bottom, try click full view, for me solved this

Anthony Taylor

Hi guy, thanks for all the work you do... Happy being a supporter! 2 questions. 1. Is there a way to get it not "load partially 5629.675.xxx etc every run? It seems it loads the same thing each time. 2. I got the Hi-dream to work with fp8 on my new sys, WOOT, can I add Framepack to it? if so what's the safest way to not wreck my currently working comfyui

Aitrepreneur

It's not blank, as said in the video, click on the fit to view icon on the bottom right

Aitrepreneur

if you are using an already existing comfyui version, once you have updated comfyui to the latest version and all the nodes, just redrag and drop the workflow inside comfyui again

bwild370

I was having the same problem! You saved me lol!!

Aitrepreneur

if you are using an already existing comfyui version, once you have updated comfyui to the latest version and all the nodes, just redrag and drop the workflow inside comfyui again and select the corresponding GGUF model for your GPU (in your case it will be the Q5 ks)

Aitrepreneur

if it's loading partially that means you don't have enough vram for that, maybe use the gguf side instead. For Framepack, it's a separate webui and it's a portable version as well so it will not affect your comfyui install either

Aitrepreneur

Guys I showed this step in the video as well, always do this when you see a blank canvas when loading a workflow no matter what

Darren Keathley

This workflow is showing blank for me. Nothing on the screen

Aitrepreneur

It's not blank, as said in the video, click on the fit to view icon on the bottom right

Midnight

Okay so this fix did not work. But reinstalling from scratch with your new bat file seems to have fixed it. No issue at all. Thank you.

Marvin Lossa

Thanks for this workflow! But there is one detail that bothers me. In GGUF I2I the fast groups bypasser for FP8 ADD DETALS IMG TO IMG cannot be deactivated. I click on it, it turns off for a moment and switches back to "yes" instantly. All other switches work.

theworldofanatnom

im using the desktop app, it dose not have an update comfyui button, it auto updates

Aitrepreneur

Woops, you're right, nice catch! Just correct and reuploaded the workflow, redownload it from this post and redrag and drop it inside comfyui 👍

Silent Monk

Missing Node Types When loading the graph, the following node types were not found ImageResizeKJv2 (In group node 'workflow>loader') workflow>loader

El Cabeleira

Just a follow up - It didn't let me git pull the KJ nodes so I deleted them entirely and git cloned them, after that it only threw 1 missing workflowloader error (had 2 before), then I installed KJ nodes nightly and dropped in the workflow again and it worked. So it was definetly something with KJ nodes most likely. Thanks for the tip/help again.

KoBa4

Please, share the site link with the artists

contrite831

Update to the latest version of comfyUI. Then in Custom Nodes manager in Comfy UI search KJnodes. Select switch ver, then pick nightly and press select. Once it updates refresh the session. Close all tabs of the HIdream workflow then re-drop the workflow into comfyUI. Hope this works for you

ti

Any help here? Prompt outputs failed validation: UnetLoaderGGUF: - Value not in list: unet_name: 'hidream-i1-dev-Q8_0.gguf' not in ['hidream-i1-dev-Q5_K_S.gguf'] PS: Can you just create a batch installer that solves the issue with the 5x series regarding PyTorch and the whole 'No kernel image is available for execution on the device' topic? That would be amazing :D

Charles Hopper

as I replied to a similar post earlier, I have even tried imputing the exact same prompt as was in the video however I keep getting this error: Prompt execution failed Prompt outputs failed validation: UnetLoaderGGUF: - Value not in list: unet_name: 'hidream-i1-dev-Q8_0.gguf' not in ['hidream-i1-dev-Q4_K_S.gguf']

Art4Ever

Thanks for the update! I am very glad it helped! Wish you all the best inspiration and creativity!

NeO

Aitrepreneur node "workflowloader" is broken on ur WF

ap0c

Thanks for sharing, much appreciated :) I find it very impressive how it can handle multiple styles out of the box with such detail. I am however very sceptical about the NSFW claims, the few tame prompts that I have tried (for testing only ofc!) were incredibly underwhelming.

Aitrepreneur

Simply updating comfyui to the latest version is enough if you have used the installer. DO NOT use the manager to update comfyui, the manager has been broken for months, to update comfyui correctly, go inside the update folder and run the update_comfyui_and_python_dependencies.bat file (do not forget to continue the install once it asks you to press any button) then, once the update is done, relaunch comfyui and re drag and drop the workflow inside.

Aitrepreneur

you need to select the model you have downloaded like it says in the workflow, by default it's the Q8 but if you have downloaded the Q4 ks you just need to click on the model node and select the available model.

Aitrepreneur

you need to select the model you have downloaded like it says in the workflow, by default it's the Q8 but if you have downloaded the Q4 ks you just need to click on the model node and select the available model.

Aitrepreneur

Simply updating comfyui to the latest version is enough if you have used the installer. DO NOT use the manager to update comfyui, the manager has been broken for months, to update comfyui correctly, go inside the update folder and run the update_comfyui_and_python_dependencies.bat file (do not forget to continue the install once it asks you to press any button) then, once the update is done, relaunch comfyui and re drag and drop the workflow inside.

Aitrepreneur

I mean...I'm not sure what prompt you used but I got full on view...you can also use loras to make it even better, that's what they are for as well

ap0c

Don't suppose you could post a couple FP8 prompts and seeds on your github so that I can see if its an issue with my setup, I used a previously configured comfyui setup, starting to wonder if that is the issue.

Charles Hopper

Thank you for helping this newbie out! I can’t believe I missed something so obvious—if it was a snake, it would’ve bit me! Seriously though, I really appreciate you taking the time to point that out. Because of your help, I’ve been able to experience some of the incredible new tech coming out for AI art, and it’s been amazing. Thanks again for your patience and guidance!

NeO

I've already done all these steps, reinstalled everything, and the error persists. The node you put in the flow doesn't work. It's easier to use another one; this specific one just doesn't work, even with the latest version correctly updated.

Aitrepreneur

It's not broken, it works, it's not that complexe, everyone who followed the steps carefully made it work, so it's on your side. You need to update comfyui correctly using the update_comfyui_and_python_dependencies.bat file until the end, use the nodes and models installer inside the comfyui folder and then redrag and drop the workflow inside comfyui. That specific node is linked to the kjnodes repo that is installed with the installer so it's nothing complicated. Follow everything i put in the comment step by step. If you have an issue you can record a video and send it to me in dm, I'll show you where things got wrong. Or you can also reinstall it from scratch using the whole installer.

Aitrepreneur

you can go on civitai and look for the uncensor lora for hidream and use the prompts on the page, this way you got lora + prompt + images

ap0c

That is what I was missing, thank you very much for taking the time to reply :) Stay awesome!

Lars Drakblod

what Loras are compatible with HiDream? and were do i find them?

.Sirene

This is where I think a lot of users are making mistakes. I closed the window after it said "Press Any Key To Continue"... I thought it was done, as is expected behaviour. In this case, guess I was too quick to jump the gun lmao. All good. Glad to see where I went wrong.

Aitrepreneur

Yes that's why I put in my instruction to continue the install once it asks you to press any key in case people quit before it's actually finished

Dat Nguyen

really dumb question, but how do we access it afterwards? like i got it to work but then it crashed. but now since its a different installer everyhings different.

.Sirene

So I did now a fresh installation, followed all steps and - did not close until the installation was truly finished. But, for some reason - the .json workflow does not load any modules. I tried a new installation, and was met with the same issue. Not sure what's the issue now, I made sure to read and double read every step before continuing.

tezromania

Is it really slow for everyone else? It is taking me 45 seconds to do generations on an RTX 5090: - euler 20 steps - 720x1280 - dev Q4 gguf model

Aitrepreneur

If you have a 5090 use the fp8 side, its much faster

tezromania

Thanks, yeah, that takes it from ~45s to ~30s. And I get about ~20s using cfg 1.0. Does that feel closer to your own experience?

Aitrepreneur

Yeah it takes around 22s to generate an image

Robb

I downloaded the FP8 and getting this Prompt outputs failed validation: QuadrupleCLIPLoader: - Value not in list: clip_name1: 'clip_l_hidream.safetensors' not in ['clip_g_hidream.safetensors', 'llama_3.1_8b_instruct_fp8_scaled.safetensors'] - Value not in list: clip_name3: 't5xxl_fp8_e4m3fn_scaled.safetensors' not in ['clip_g_hidream.safetensors', 'llama_3.1_8b_instruct_fp8_scaled.safetensors'] VAELoader: - Value not in list: vae_name: 'ae.safetensors' not in ['taesd', 'taesdxl', 'taesd3', 'taef1']

Aitrepreneur

were there any issues during the install of the models and nodes? then have you updated comfyui using the update_comfyui_and_python_dependencies.bat file?

Joshua Kerns

So when I am going though the update comfyui_and_python. I get to the point where it ask me to press any key to continue and when I do it closes the window. I assume it not updating. Any idea what could cause that?

Alex20 Syn

This isn't a technical question but a stylistic one. It kind of looks like HiDream has the character staring at the camera problem. One of the things that made Flux so powerful was that poses were more varied. An image detail can be massaged with a number of techniques, but if the pose is not dynamic you are kind of stuck. How is HiDream at more candid poses?

theworldofanatnom

idk why he doesn't mention this but if you still cant get the nodes to work and not using the installer: use the installer, When you run it put it in a empty folder were u can drag the nodes and the models to were they need to be. u can also just stop the install after the nodese finish, here are some nodes that the comfy manager isn't detecting for some reason.. Aitrepreneur i love you and your work man but u really gotta understand that alot of the stuff you take as common sense is complex to some of your fans especially if their in unique scenarios like the desktop app which dose not have a update_comfyui_(i forget this part).bat but you keep telling them to click it regardless.

Bicklemode

Thank you your reply and our king Airepreneur's reply fixed it for me!

Silent Monk

How would I do this inside the Pinokio version of ComfyUI?

Shivam Singhal

its giving me a blank workflow when I add the json

Soham Dasgupta

I had high hopes, but the image-to-image outputs are quite underwhelming. I'm using the full model with everything at the highest settings on a Mac, and sometimes the images come out without limbs. I don't know if it's because I'm on a Mac.

Frunk

Like many others I went through the process and it just came up with a blank canvas... before realising I needed to update first. I DID get it working but I had to cross a few bridges. Here is what I did. Skip the bits that already work fine for you. ----------- I followed the instructions and failed to get the updates working as highlighted in the original post. Just kept getting these errors: Traceback (most recent call last):   File "E:\ComfyUI_windows_portable\update\update.py", line 66, in     ref = repo.lookup_reference('refs/remotes/origin/master')           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ KeyError: 'refs/remotes/origin/master' This is because git has NEVER been run locally against the repo so it is lost. That was an easy fix... From the ComfyUI folder run: git remote -v git fetch origin That lets me run the update file "update_comfyui_and_python_dependencies.bat" as described above. Then I restarted ComfyUI using "run_nvidia_gpu.bat" I dragged in "HIDREAM_EASY-WORKFLOW.json"... It STILL came up blank. At this point I was about to give up. However on quitting the current ComfyUI instance and just restarting it AGAIN - it all burst into life and works as shown in the video. It remembered I had dragged in the JSON file but this time the UI was visible on the canvas. A few extra steps that a "one click" but I thought I would let you know how I eventually got success. Try the updates a couple of times, try dragging in the JSON file and do the odd restart of ComfyUI

Diego Vargas Escobar

Hi @Aitrepreneur, one Q, I saw in some foros that having Triton helps with the overall speed. Does the auto installer do this? thanks

Dat Nguyen

what? i ran the fresh install, no such files exist... thats why its odd....

SuperSpamTube

Its probably not centered right. Click the "fit view" symbol on the bottom right. Its a frame icon.

Inkdazzle (edited)

Comment edits

2026-09-08 20:00 the attached json is not at all as described. It is very simple. Please give us the right file
2025-05-07 17:46

Aitrepreneur

Because my installers and files and tutorials are based on the portable version (which for me is the best version since you can use it anywhere with not specific requirements and move folders freely)

Aitrepreneur

I have a special installer for that but I'm keeping it for future video, still need to test it on a 5000 series gpu

Bill P

-------- Downloading ComfyUI -------- % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0 43 1974M 43 850M 0 0 355k 0 1:34:41 0:40:45 0:53:56 421k curl: (18) end of response with 1179463021 bytes missing Download failed. Press any key to continue . . . Terminate batch job (Y/N)?

Steve Hanff

same issue, Comfy UI updated today, updated all nodes but that one is missing

Will the Thrill

This helped me with my issue as well! Thank you!

Aitrepreneur

I really don't understand what you are saying, dm me screenshots of your screen

Steph Willis

When i add the json file there are no nodes at all its just black, i cant see any errors kicking off in the console behind :| I just ran a full installer using the bat file didnt notice any errors

Steph Willis

Ignore me clicking the center view button made it work :)

steamrick

I can confirm, the workflow loads not centered and looks empty

Alex Litvin

Thank you!! I was wasting WAY too much time trying to figure out what was wrong. lol

Chroma

Yeah the JSON file doesn't work nothing appears when you drag and drop into ComfyUI

Aitrepreneur

As showed in the video, click on the fit to view icon on the bottom right

4rianne

Hiii, Im trying to make it work on an AMD gpu with 6gb vram, any advice or tips? Im installing HIP SDK and I think this method uses Zluda.

Chroma

Ohh k I'm just special thank you 😂🤦‍♂️

Aitrepreneur

For amd you need the special zluda comfyui install

Shivam Singhal

this cant produce nsfw stuff? I have tried and it doesnt seems to work

Aitrepreneur

depends what I suppose, you can also go on civitai and look for specific hidream lora for that and take inspiration from the prompt used on that lora page

Niklāvs Trankalis

Where can I find the workflowhiinpai nodetype? It's the only Missing Node Type.. everything else seems to be fine

Morpheus

Thank you that solved it!

Aitrepreneur

then that means the initial node was not installed correctly previously. Go inside the ComfyUI_windows_portable\ComfyUI\custom_nodes folder and delete the ComfyUI-KJNodes folder. Then while still inside the custom nodes folder, click on the folder path, type cmd press enter, this will bring a command prompt window inside that folder and inside that cmd window type git clone https://github.com/kijai/ComfyUI-KJNodes.git ..\..\python_embeded\python.exe -m pip install -r ComfyUI-KJNodes\requirements.txt press enter and wait until everything is installed. Then relaunch comfyui and redrag and drop the workflow inside comfyui again.

Bobby Boe

Hi my friend. I tried to use your workflow, but one node "workflowloader" refuses to get installed. I wonder why. I have Hidream in another workflow running, but was curious about your workflow. Is that "workflowloader" essential? could it be substituted? I am running the desktop-version of ComfyUI, in which it is a little complicated (for me) to install via pip... . so usually I try to do everything using the manager. (yes, I updated everything). What package does the missing node "workflowloader" belong to? I am using 4060TI 16GB, so I would use the right GGUF-side. ------------------------ Actually now, I saw, this node is the image-loader in your img2img-workflow. In the worflow the name was "workflowloader" (and red because missing), and in your YT-Video, it is only called "loader". I can see that is has some more options than a normal image-loader. Can you tell me what package it is from, and if the additional options are really needed (and why?)

Paul Mazurkiewicz

Hey AIT, I tried substituting in the full model instead of the DEV model (Quant 8) and it failed...just produces garbage. What else do I have to change in the text to image GGUP or FP8 workflows to use the full model instead of the DEV model? Is there a difference in quality?

Onex Sys

After installing the Hidream "Update" to a prior installation of Flux with using your 3.1 Workflow - Suddenly the 3.1 Workflow gives me OOM exceptions when creating from text. Image to Image still works fine. I tried a completely new installation into a new folder of the "old" 3.1 workflow, but It leaves me with missing nodes (tried all your fixes from troubleshoot video) but still have those errors: workflowFlorence workflowFlux Controlnet workflowNodes 3 workflowXLabs Sampler workflowAdetailer workflowoutpaintflor workflowfaceswap2 workflowfaceswap3 workflowREACTORFACESWAP workflowpullflo workflowpulidtxt workflowpluidlux workflowfillparam workflowPULIDREDUX workflowREDUXOTHERPARAM workflowshaker workflowLiveparam workflowLIVEPORTRAITIMGPARAM workflowLIVEPORTRAITVID2VIDPARAM workflowLTXPARAMTEXTOVIDGGUF workflowLTXPARAMTXTFP workflowLTXIMAGETOV workflowhuynyav2vparm do you have any hint for how to resolve either? Happy with either fixing the combined comfy flux + hidream or just having the flux only restored (since I find the flux results superior to hidream)

chucka

Yip same here, hope this can be fixed

KOTNO Universe

I get an empty canvas error when I click on the fit to view icon on the bottom right.

lurkergao

Hey I downloaded it and did a fresh install and it works great but the image to image is broken on my install. says the following: Prompt execution failed Prompt outputs failed validation: ImageResizeKJv2: - Value not in list: device: '2' not in ['cpu', 'gpu'] every run across this error before? any idea how to fix it? it's just effecting the image to image

Raevn Eldaer

The workflow is just empty when I load it...

Aitrepreneur

click on the fit to view icon on the bottom right

lurkergao

I figured this out for anybody else that gets this error. in the workflow device option switch it to cpu or gpu

Omar Abouzaid

Do you ever plan to do a HiDream LoRA training guide video? You mentioned some interesting points in your video such as the fact that the dev model is better and that the provided settings were no optimal for training and I'd love to hear some more details about that as well as to know which training client you prefer (ai-toolkit, diffusion pipe, etc.)

Starbuckk Serapis

This is producing censored images. Am I mssing a step?

Onyx N

the setting is messed up every time i restart th comfyui, and always running with cpu eventhough i already choose gpu

Onyx N

i think this workflow really need an update, its censored

Vincent Nestler

I am unable to find the node - Upscale by Factor with Model (WLSH) Thoughts anyone?