1-Click INSTALL FLUX & ComfyUI WebUI!

Look for the ones on CivitAI for FLUX, there are several good realistic and otherwise models for Flux on that site.

HealingPaint

You are a bit late to the party on this one! I've been using Flux in Comfy for many months...was this due to your medical-related absence? ....understandable; it's good that you created this for those who haven't checked it out yet.

HealingPaint

Thanks for this! I'm wondering how I can generate more realistic, low-quality images, like I've seen others achieve by using prompts such as "IMG.0123.CR2" to create that realistic vibe. When I try this with my setup, it doesn't work and instead produces images that are too high quality. Or ss this flow not designed for that?

Floris

Thank you for the method! May I ask you a checkpoint to obtain super realistic portraits?

Mauro Prelli

Instead of running the bat file, I ran it from a cmd window so it would stay open and I could see the errors. (I renamed the .bat file to a smaller file name for easy of use in cmd) The first one I got was that 7z was not available, so I installed it. Then I got the error that ComfyUI_windows_portable_nvidia_cu121_or_cpu.7z could not be found. So I went to the manual instructions version, got the link for the file, renamed it to match "ComfyUI_windows_portable_nvidia_cu121_or_cpu.7z" exactly then ran it from the cmd again. At this point I'm not 100% sure it's working, but I did at least get into comfy.

Baubleheadz

Hello. Thanks for these workflows. I was very excited to try them out and only got black images for a while. Until I set the t5xxl_fp8_e4m3fn.safetensors model for clip 2. Maybe that's helpful to someone else who overlooked that little detail. @Aitrepreneur can you maybe explain a bit why "Load a workflow and make sure that you are using the t5xxl_fp8_e4m3fn.safetensors model for clip 2 so that you can begin generating your images" is important. It would also be cool if you could provide a version on Lightning AI. They give away a bunch of free hours, and it's easy to replicate studios.

Edgar Hoover

Couldn't find the json you mentioned in the video for the muitple A1111-like routines, i found 01-FLUX-FAST-LOWVRAM and a few others here.

K

it's a bat file, you can look inside

Aitrepreneur

Hey people if you get errors eg weight not right or other strange errors with the nodes, before you go trouble shoot, right click on the node and select "reload" not all .json files are correctly compiled when saved. It can save you a tone of grief if you try that first. Other errors like it can't find certain, files. Just because XLABS had to be different, they wanted their models to be in a specific place and folder, it wasn't good enough for them to have their models in the standard comfyui models folder. You have to create a models folder in the root of your comfyui install folder, named Models, then xlabs then have the following folders in there controlnets, flux, ipadapters, loras ... Other things you can check for is sometimes the creators of .json files have their models in subfolders, eg Models/controlnet/this is where i want you so I can separate you from the rest folder... Comfyui is smart and generally will pick up subfolders in models folders, sometimes its not so smart, so you may need to look for where your actual models lives. If you use the extra_model_paths.yaml not all models will be picked up from there in particular anything to do with XLabs. Hope this helps some of you out. Its never ever a one touch for everyone! Sometimes you just need to work at it.

grumpy grumpy

i click on add lora button but nothing happens?

Silly Pear

heres another one for you : your .bat is not calling run nvidia.... properly. [INFO] Running ComfyUI... '"run_nvidia_gpu.bat"' is not recognized as an internal or external command, operable program or batch file.

grumpy grumpy

Ok, you need to work a little on your .bat file, reason why. Not everyone installs comfyui in the default folder name comfyui portable or whatever... I have mine installed in a dir called ComfyuiStandalone... running your .bat I get the following errors.. ""The system cannot find the path specified."" for every file it is downloading. For me not such a big deal once I found where it was dumping the files, in my D: root ... Your .bat doesn't check for an existing Comfyui install. Just assuming the dir is a default name isn't particularly wise. Once all the downloading has finished Ill just pull the files to where they need to go, I use a .yaml I put all my models in a universal dir and serve it out to whatever needs them using symlinks.

grumpy grumpy

Type: Trojan Name: HEUR:Trojan.BAT.Agent.gen Precision: Heuristic Analysis Threat level: High Object type: File Object name: ULTIMATE-TTS_AUTO_INSTALLER.bat Object path: D:\AI models\Ultimate Flux Workflow\text to speach MD5 of an object: BF72EF2AF9A2AC9C3A2926367B5EDA61

grumpy grumpy

read the first line in the post update + load workflow

Aitrepreneur

I'm getting this exact same error.

John Odum

keep getting error saying ive loaded workflow and updated but still get this Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3']

koozy

Sorry, where are these customer installer bat files located? Can find any active links. Nvm - i see them down the list.

Michael

any error message?

Aitrepreneur

looks like the clip_l.safetensors file wasn't downloaded properly, try relaunching the installer in a new folder or dl the file with this link: https://huggingface.co/Aitrepreneur/FLX/resolve/main/clip_l.safetensors and place it in the comfyui/models/clip folder

Aitrepreneur

Same for me also and i have 7zip so don't think that is it.

LexisGecko

I ran FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat I selected both A options. But it doesn't install. Do I need 7zip installed? I currently use WinZip.

Randal Woody

It looks like something didn't load properly, I restarted everything and still get that error but it runs fine?

Tim Eilers

and I did run the update

Tim Eilers

I am getting this error message- Prompt outputs failed validation DualCLIPLoader: - Value not in list: clip_name1: 'clip_l.safetensors' not in ['t5xxl_fp16.safetensors', 't5xxl_fp8_e4m3fn.safetensors'] When I ran the bat, I did notice that it gave me an error on one of the items from huggingface, is a file missing?

Tim Eilers

read the first line in the post update + load workflow

Aitrepreneur

I keep getting this error "'conv_in.weight'" any ideas?

ker0

read the first line in the post man :) update + load workflow

Aitrepreneur

I have a clean install of Windows 11 and ran FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat and then FLUX_AUTO_MODELS_INSTALL.bat but I'm having errors in all the json that I try to execute, here's an example: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] Python 3.10.11

Demetrius Ricarte

on this page, in this post, it's there

Aitrepreneur

try another browser maybe?

Aitrepreneur

yes, I'll show how in the next update coming in a few days

Aitrepreneur

Where is the FLUX_AUTO_MODELS_INSTALL.bat file?

Daniel Le Blanc

So I followed your instructions on this one and get to the point we're you update comfyui thru the manager and clise/reopen browser. After that the webpage comes up but is blank. I was using option b code. Have you seen this before? Is there a way for me to go in there and restart the web interface manually ?

Daniel A Perez

can you do inpainting from this somehow

AR

It says there is a security error when I do.

Wally Hunt

yeah makes sense, a mobile 3050 ti has like 4GB of vram. You won't be able to run Flux with that, maybe a quantized model (I'll show it in the next update video)

Aitrepreneur

Laptop 3050 TI - Low-Vram runtime with matching json. Already updated ComfyUI per the post. Getting this message and then it just sits. .. loaded completely 0.0 9319.23095703125 True clip missing: ['text_projection.weight'] .. just trying to queue the 01-FLUX-FAST-LOWVRAM.json workflow with no edits.

Zolon Farsane

you should be able to install custom nodes as is, can't you?

Aitrepreneur

yes, everything new is coming this week

Aitrepreneur

On Runpod, how do I enable weak security settings so I install custom nodes? When I enable it in the config ini file and restart I still cannot install custom nodes.

Wally Hunt

Do you have any LoRA workflows? How about upscalers?

Joseph Raegan

read the first line in the post, update the webui and load the workflow in the post

Aitrepreneur

Hi i keep getting this error when i try to queue prompt: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] Can anyone help me?

Lars Boelaerts

again that means you are not using the correct workflow then, send me dm screenshots of the webui

Aitrepreneur

yeah looks like they deleted the file from the repo and replaced it with a safetensors file instead, I'll be updating the installers soon anyway for the next video

Aitrepreneur

That didn't work, it didn't download the missing checkpoint file. From manually downloading and placing v1-5-pruned-emaonly.ckpt file, it doesn't seem to be Flux. It seems to be Stable Diffusion 1.5. I compared the same input with an online Flux generator and my local output sucks like old SD - messed up hands, sometimes messed up faces and the occasional anatomical nightmare fuel.

Gator3D

maybe you wanna check if the curl command is still up to date. I now instead used wget --content-disposition https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors to download the file, and renamed it to flux1-schnell.sft everythings working now. :) Anyways, thanks a lot for all your awesome work! <3

Jens

I have deployed it on Runpod but the flux1-schnell.sft is not downloading correctly. the following sh-prompt only downloads a 15 Byte sized file, content "Entry not found": curl -L -o "flux1-schnell.sft" "https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.sft?download=true"

Jens

I had to, and did the exact reverse: it was sending me tro 127... and for some reason, my rig never connects to the IP. I need these things to send me to localhost:port. I just edited the main.py file making the change. Worked a charm. I feel stupid for not having realized that was the issue immediately, but in my defense, I've been quite overworked the past month and just wasn't focusing enough, never having thought to eyeball the address bar. It is what it is. Thanks for all you do, Q.

JD Turner

Usually comfyui will automatically send you to: http://127.0.0.1:8188 you can either just copy and paste that url or just put that into your favorites if for some reason it doesn't open the window automatically

Aitrepreneur

no need for that, just read the first line in the post, update the webui and load the workflow in the post.

Aitrepreneur

Yes very veeeeery soon

Aitrepreneur

OK, on a hunch ... the problem was that the address autolaunch was sending me to was 127.0.0.1 instead of localhost. I've had connectivity issues with 127.0.0.1 with a couple of other gradio interfaces before, but getting an error message and tracebacks to chase down was a first for me this time around. Hopefully my being stubborn on giving up will be helpful to someone else experiencing this issue.

JD Turner

I've scoured the googles and am not finding anything helpful. When I attempt to connect to the GUI, instead of the GUI I get this: Got more than 8190 bytes (14396) when reading Header value is too long. Searches reveal that this is a somewhat common problem in a plethora of things that aren't ComfyUI, but nothing specific to this is returning in my searches. Any tips or leads would be greatly appreciated.

JD Turner

Hello. First thanks for your content. "IF YOU GET A RED ERROR IN THE UI MAKE SURE YOU HAVE A MODEL/CHECKPOINT IN: ComfyUI\models\checkpoints You can download the stable diffusion 1.5 one from: https://huggingface.co/runwayml/stable-diffusion-v1-5/blob/main/v1-5-pruned-emaonly.ckpt" I can't get the file. Do you have another link please?

Bruno

I have the same problem Ive updated it, Prompt outputs failed validation UNETLoader: - Required input is missing: weight_dtype

Raevn Eldaer

Will you provide a workflow for checkpoints from Civitai and Lora models? In a way that is connected to a good upscaler? For me and other beginners with Comfy, this would be a great help.

Tomasz

Now it says: C:\Users\Brandon\a1111\ComfyUI_windows_portable\New folder>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build The system cannot find the path specified.

Otaku_Hedgehog

Apologies i only watched the video, it wasnt clear in the video where to get the json. Is there a model for comfy you recommend for cpu?

Ma No

read the first line in the post, update the webui and load the workflow in the post.

Aitrepreneur

read the first line in the post, update the webui and load the workflow in the post. However if you don't have a GPU you'll have trouble running a model like Flux correctly

Aitrepreneur

Also do you have the workflow?

Ma No

I'm missing the checkpoint file, but the link in the readme is broken and no longer appears to be available on huggingface. What checkpoint file can I use? Also why stable diffusion v 1.5?

Gator3D

No Nivida GPU so runing the CPU - this is the error i get Prompt outputs failed validation CheckpointLoaderSimple: - Required input is missing: ckpt_name

Ma No

Well, that was mildly embarrassing, to have to admit to not having read the first line of the instructions. Thanks so much. I will say, the first prompt (as a test) took 694.26 seconds to execute, although the second effort took only 31.89 seconds. Maybe it just has trouble getting started.

THEO-Ed animation

have you updated comfyui first? also are you sure that the workflow has been loaded correctly?

Aitrepreneur

Try relaunching the installer in a new folder and check if there is no error, might have been an issue during the download

Aitrepreneur

I am on a lowly 1070 but I want to get slightly familiar with how this works before I stat shelling out money for Runpod. After installing I run the 'run_nvidia_gpu-LOWVRAM'and then load the 01-FLUX-FAST-LOWVRAM... I do not get the Load Diffusion Model box, in its place I get a UNETLoader box which has the part to select the unet-name but not the weight_dtype. that option is not present.

Jason Hancox

Hello can I get some help please? Everytime I try to launch the run_nvidia_gpu.bat file, I get the following error: "OSError: [WinError 126] The specified module could not be found. Error loading "C:\Users\Brandon\a1111\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\lib\fbgemm.dll" or one of its dependencies." I currently have an RTX 3090 and I already updated ComfyUI in the update folder, but the program still refuses to launch?

Otaku_Hedgehog

yes read the first line in the patreon post, update comfyui + load the workflow

Aitrepreneur

try in another browser and click refresh

Aitrepreneur

try in another browser and click refresh

Aitrepreneur

Okay, I've got this up and running - sort of (ComfyUI looks so much different from Automatic1111) - but I can't actually get the program to generate anything, because I keep getting the following error messages: in VAELoader, it tells me "Value not in list: 'ae.sft' not in ['ae.safetensors', 'taesd', 'taesdxl', 'taesd3']" Which is weird, as I do have ae.sft in the ComfyUI/models/vae folder (and I don't have any of the "taesd" files it talks about that I know of) in DualCLIPLoader, it reports "Value not in list: type: 'flux' not in ['sdxl', 'sd3']" Again, I have flux1-dev, flux1-dev-fp8 and flux1-schnell-fp8 in the ComfyUI/models/unet directory, and have no idea why it's not seeing any of them, or where it sees sdxl or sd3 Any idea as to what I might be doing wrong?

THEO-Ed animation

I also got a white screen after updating comfyui. Any update on this situation?

Russ Ader

Happen to me too after update Comfyui

Camilo Osorio

I have just a white screen when opening comfy ui

Kristopher Dean

Thanks works like a treat :)

Kris Bennett

Update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column then load a 01 json workflow file

Aitrepreneur

Finally got around to the point of actually running this and get this message: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] UNETLoader: - Value not in list: unet_name: 'flux1-dev.sft' not in ['flux1-dev-fp8.safetensors', 'flux1-schnell-fp8.safetensors'] How can I fix this @5:17 the video just states I need to drag that file in and queue a prompt after installing everything with the bat file as per the video. Thanks, K

Kris Bennett

Can this link be manually inputted into the auto installer or could you provide instructions how to install ComfyUI?? I imagine once comfyUI is installed then I can use the Flux Installer

Otniel Ocampo

Soon™

Aitrepreneur

soon?

NorthDtagg

Unfortunately that didn't help.

Camelot VFX

I'm not quite sure. I noticed that in the Update folder you had a bat that updates python dependencies as well as the UI itself. Maybe I should try that?

Camelot VFX

well I don't use linux so I can't really make sure that it works, altough the runpod ones are probably what's closest to a linux installer

Aitrepreneur

Maybe because in your case you are running windows in a VM so it can't correctly find the git install

Aitrepreneur

yup that could be it, what's your python version?

Aitrepreneur

Thank you for this video and the installers! It would be nice to see local Linux one-click installs in the mix along with the cloud-based runpod .sh files.

Nimodes

Ok ... I modified your bat to use "where git" and check to see if the %ERRORLEVEL% = 1 (means found in the case of where) before trying to run it and that seemed to work - not sure why the 9009 error seems to be un-resetable!

Kris Bennett

I have an issue when running your FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat it fails to install git.. Basically if you don't have git installed the version check sets %ERRORLEVEL% = 9009 and for some reason this seems to be able to be set again by following commands - meaning that your with your second check against it, it is still 9009... Not sure how this can happen - I just installed it manually.. This was attempting to install it in a clean install of Windows 11 in a VM btw.

Kris Bennett

Oh, I am seeing this in the cmd window. Not sure if I've done something wrong, or if this matters: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.)

Camelot VFX

Literally 40 minutes, believe it or not. I'll try the fp8 instead, and will try a new folder to see if it makes a difference. It also seems like it shouldn't need to unload and then reload the model every time I change the prompt, right? Yeah this is very weird. I appreciate you trying to help though.

Camelot VFX

40 minutes....??? Is that an exaggeration or the actual amount of time it takes? Because if it is, there is seriously something wrong here and I'm really not sure what...have you tried the t5xxl fp8 version instead of the fp16? With your gpu it should load a bit quicker...but 40 mins is insane...especially with such a powerful rig... have you tried reinstalling in a new folder, see if that makes a difference somehow?

Aitrepreneur

No error in the cmd window, yes I'm using the 01 json workflow. No error during install. It seems to work okay (I could get the test image of a lion like you did in the video), but it takes like 40 minutes to load the model first and then it generates the image in about 30 seconds. And if I change the seed on that same prompt, it's very fast too. But the moment I change the prompt to something completely different, I can see in the command window that it unloads the model and then starts loading it again, and it takes another 40 minutes.

Camelot VFX

definitely not, do you have any error in the cmd window? Are you using the 01 json workflow? Any error during install? Have you tried reinstalling it in a new folder as well?

Aitrepreneur

yes

Aitrepreneur

How do i uninstall it? is removing the folder 'ComfyUI_windows_portable' enough?

User#5c2e504f

It also seems like it unloads and then loads the model every time I change the prompt...

Camelot VFX

Hey there, I did the "Local Install if No comyfui is installed" and updated comfy, used the json, everything seems fine. But when I try queueing the image, I can see in the command prompt that it's loading the model, and it seems to take forever. I have a 4090 and 64 gb of system ram. Should it be taking 20+minutes to load the model?

Camelot VFX

yup it's a brand new model, I will talk about it in my next video coming Soon™

Aitrepreneur

yes, it will generate images much faster with no quality difference

Aitrepreneur

you need to use the special comfyui for AMD gpus: https://github.com/patientx/ComfyUI-Zluda

Aitrepreneur

I had so much trouble with the ones in the video here. I found a NF4 version of the model and it took my times from 14 minutes to 7 seconds on a 3080 with 10gb VRAM

Sam Shennan

We should pick Fast-Lowram install even with a 4090?

eiyukabe

ahaa, i did not have git installed, problem solved :D , also i disabled antivirus on windows

Anders

Is there something I can do to make the program run with AMD GPU's?

Otniel Ocampo

Hi been trying to install the ComfyUI-Zluda and thought I had got it sorted, but just can't get it to run after multiple attempts, I keep getting as far as the following; C:\Users\MaxPh\ComfyUI-Zluda>C:\Users\MaxPh\ComfyUI-Zluda\start.bat *** Checking and updating to new version if possible Already up to date. [START] Security scan [DONE] Security scan ## ComfyUI-Manager: installing dependencies done. ** ComfyUI startup time: 2024-08-18 20:36:38.503017 ** 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: C:\Users\MaxPh\ComfyUI-Zluda\venv\Scripts\python.exe ** ComfyUI Path: C:\Users\MaxPh\ComfyUI-Zluda ** Log path: C:\Users\MaxPh\ComfyUI-Zluda\comfyui.log Prestartup times for custom nodes: 1.5 seconds: C:\Users\MaxPh\ComfyUI-Zluda\custom_nodes\ComfyUI-Manager Traceback (most recent call last): File "C:\Users\MaxPh\ComfyUI-Zluda\main.py", line 83, in import comfy.utils File "C:\Users\MaxPh\ComfyUI-Zluda\comfy\utils.py", line 20, in import torch File "C:\Users\MaxPh\ComfyUI-Zluda\venv\Lib\site-packages\torch\__init__.py", line 141, in raise err OSError: [WinError 126] The specified module could not be found. Error loading "C:\Users\MaxPh\ComfyUI-Zluda\venv\Lib\site-packages\torch\lib\caffe2_nvrtc.dll" or one of its dependencies. C:\Users\MaxPh\ComfyUI-Zluda> I have gone to the folder and can confirm the so called missing file is in fact in the folder and is 16kb in size. I have tried re-installing, patching, ect.. and always get to the same point, any assistance would be most appreciated. Quick note, the PC spec is; AMD Ryzen 9 7950X3D 16-Core Processor 4.20 GHz Installed RAM 64.0 GB (63.1 GB usable) 64-bit operating system, x64-based processor Windows 11 Pro AMD Radeon RX 7900 XTX 24gb graphics card Edit; Just tried reinstalling Python due the the following comment in the readme: Problems with "caffe2_nvrtc.dll" : if you are sure you properly installed hip and can see it on path, please DON'T use python from windows store, use the link provided or 3.11 from the official website I did install from the Python website but it was version 3.11.9 so I uninstalled this and instead installed version 3.10.11 as this was first option given in the notes. I then deleted the venv folder and re-ran the install.bat. After doing all of this I still have the same error when trying to run, and I can't think what else to try at this point. So really do need some help guys and girls, my software skills are ancient to say the least, last attempt to run below; C:\Users\MaxPh\ComfyUI-Zluda>start.bat *** Checking and updating to new version if possible Already up to date. [START] Security scan [DONE] Security scan ## ComfyUI-Manager: installing dependencies done. ** ComfyUI startup time: 2024-08-18 23:43:22.591863 ** Platform: Windows ** Python version: 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)] ** Python executable: C:\Users\MaxPh\ComfyUI-Zluda\venv\Scripts\python.exe ** ComfyUI Path: C:\Users\MaxPh\ComfyUI-Zluda ** Log path: C:\Users\MaxPh\ComfyUI-Zluda\comfyui.log Prestartup times for custom nodes: 1.5 seconds: C:\Users\MaxPh\ComfyUI-Zluda\custom_nodes\ComfyUI-Manager Traceback (most recent call last): File "C:\Users\MaxPh\ComfyUI-Zluda\main.py", line 83, in import comfy.utils File "C:\Users\MaxPh\ComfyUI-Zluda\comfy\utils.py", line 20, in import torch File "C:\Users\MaxPh\ComfyUI-Zluda\venv\lib\site-packages\torch\__init__.py", line 141, in raise err OSError: [WinError 126] The specified module could not be found. Error loading "C:\Users\MaxPh\ComfyUI-Zluda\venv\lib\site-packages\torch\lib\caffe2_nvrtc.dll" or one of its dependencies. C:\Users\MaxPh\ComfyUI-Zluda>

Max Ryan

crashing? as in the cmd closes automatically? Try this first: open a command prompt window and drag and drop the installer file inside the window, then press enter, this should avoid the window from closing and it will at least give us an error code we can use to troubleshoot the error, make also sure that you have disabled any antivirus/firewall from running and blocking the file/connection.

Aitrepreneur

FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat, crashing, and the other link dosent exist anymore, can we get a new link ?

Anders

Hi, it took a while using your instructions to install LOCAL INSTALL If NO comfyUI is installed : directions on my old computer as a test to see if I could do it, and happy days finally got it up and running. Any chance on a step by step on installing this AMD version? Followed the link, I assume I install everything as per that link and then come back here to follow the LOCAL INSTALL If comfyUI IS INSTALLED : section for flux. I'm not great with software and I'm in need of a little guidance, Thanks in advance.

Max Ryan

As shown in the video, you need to change the fp16 version to fp8, on runpod only the fp8 version is downloaded since it's the fastest

Aitrepreneur

ok i updated it from th UI and that work's

Mister zede46

Hello ! he keep saying to me Prompt outputs failed validation DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] and i don't know if it's normal but i have a node "UNETLoader" instead of "Load Diffusion Model" and i have only one entry on it

Mister zede46

I installed it via RunPod, and it gave me this error when using Workflow: 01-FLUX-FAST-LOWVRAM.json . Prompt outputs failed validation DualCLIPLoader: - Value not in list: clip_name2: 't5xxl_fp16.safetensors' not in ['clip_l.safetensors', 't5xxl_fp8_e4m3fn.safetensors'] . What could it be?

Renato Vilarouca

it's installed wherever you ran the installer from, unless you ran as admin (which you shouldn't do) in that case it's installed in system32

Aitrepreneur

apparently they already did, I'll be making a video training soon, the video after my next one

Aitrepreneur

Hi there!, thanks for this!. Wuick Questions. Where is this installed to? can i change the installation folder?

User#5c2e504f

I was using the one on normal page. And it work, ~16 min to generate one 512x512 image on D version. I have 8gb bram only.

Radosław Błesiński

Will you be waiting for kohya to catch up before you make a training video?

Virtamouse

You can try this one: https://we.tl/t-01Iq68HQGb

Aitrepreneur

Hi all, I'm completely new to installing local AI, and I'm not very familiar with the mentioned certificates. Can you tell me what I need to complete the installation of ComfyUI? Here’s the image. The CMD crashed, so I repeated the process and took a screenshot before it closed again. https://ibb.co/rGFjqgx PC specs: Win 11 64bit/64gb DDR4 ram/3080Ti 12gb/Ryzen 3600

Nenad

even with the one I just gave you?

Aitrepreneur

Ok, I forgot to mention, I instal it, but after thrid opening.

Radosław Błesiński

this error is rare but can happen for some reason, try to update comfui as detailed in the post

Aitrepreneur

You can try this one: https://we.tl/t-01Iq68HQGb

Aitrepreneur

looks like some antivirus issue but it could be a multitude of reasons, first open a command prompt window and drag and drop the installer file inside the window, then press enter, this should avoid the window from closing and it will at least give us an error code we can use to troubleshoot the error, make also sure that you have disabled any antivirus/firewall from running and blocking the file/connection You can also try this one: https://we.tl/t-01Iq68HQGb

Aitrepreneur

An error occurred while installing the FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat file and the installation process was interrupted, what could be the problem? OSError: [WinError 126] The specified module could not be found. Error loading "d:\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\lib\fbgemm.dll" or one of its dependencies.

Taras

Full installer immediately closes for me trying to install on a secondary drive sadly.

Asbu

When trying install from file you provided, well one word come to mind: Guano. When it reach around 80% complete it turns off and I have to do over.

Radosław Błesiński

that was it. Probably should read the instructions .... In my defense it was one of those weeks at work, and this one is also starting to be one of those weeks. :D

Will Simon

Yes there is you need to use comfyui for AMD gpus instead: https://github.com/patientx/ComfyUI-Zluda

Aitrepreneur

I'll soon make a video explaining that

Aitrepreneur

update comfyui and load the json file workflow as explained in the patreon post

Aitrepreneur

update comfyui and load the json file workflow as explained in the patreon post

Aitrepreneur

definitely use the t5xxxl-fp8 for faster inference with that kind of GPU. For the list of prompt it all depends on what kind of prompts you wanna use, you can either use dynamic prompts or the load from file workflow: https://www.andreszsogon.com/comfyui-load-prompts-from-text-file-sample-workflow/ for the output name, you can follow this simple guide as well: https://www.andreszsogon.com/change-output-file-names-in-comfyui/

Aitrepreneur

HI both, thanks for the help, up and running now, missed the json file part first time round. All I need now is a nvidia graphics card, both my computers have Radeon cards, unless there is a way to use these cards instead. With CPU only an image takes 2 hours.

Max Ryan

How can I use negative prompts and cfg in that workflow? is there a way to use keyword blending like automatic1111?

Indrid Cold

nope. same error

Will Simon

Trying running the model download script from D:\ComfyUI_windows_portable\ComfyUI now to see if it fixes. Originally I ran it in D:\ComfyUI_windows_portable

Will Simon

Error occurred when executing CheckpointLoaderSimple: ERROR: Could not detect model type of: D:\ComfyUI_windows_portable\ComfyUI\models\checkpoints\flux1-schnell-fp8.safetensors File "D:\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_portable\ComfyUI\nodes.py", line 518, in load_checkpoint out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, embedding_directory=folder_paths.get_folder_paths("embeddings")) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 513, in load_checkpoint_guess_config raise RuntimeError("ERROR: Could not detect model type of: {}".format(ckpt_path))

Will Simon

I can't seem to get the run file to load flux. I got the files from the auto install but can't seem to get the run file. Help!

Jesús

I wonder if that's what is bottling me, because I have 16gb. Thanks for your reply

Jason O

i have 12gb vram and 32gb "normal" ram and it takes me like 150 sec at 1024x1024 with the dev model

Terr0rClown

Thank you so much for clearing that up, you are correct and I had not considered that - f**king Windows! :D

Frank B

GomezBro

I just want to make sure i'm not doing something wrong. I was able to run Flux in low vram mode using the 20 step flux1 model and it took 26 minutes. The flux1 schnell 4 step took about 6 minutes. is this expected with 12gb vram at 1024x1024?

Jason O

thanks i did downloaded it now and wil give it a try

rafik eskander

It should have been downloaded during the comfyui install, otherwise you can dl it here: https://we.tl/t-8Rb9JzxKHe

Aitrepreneur

I too do not have the run_nvidia_gpu-LOWVRAM.bat - where can it be found, google was little help.

Jason Hancox

I did now it is running but very slow as i have a 8 GB and also i don't have this run_nvidia_gpu-LOWVRAM.bat

rafik eskander

have you updated comfyui and loaded the json like I explained in the patreon post?

Aitrepreneur

Hi I am getting this error after doin the installation Error occurred when executing UNETLoader: 'conv_in.weight' File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\nodes.py", line 829, in load_unet model = comfy.sd.load_unet(unet_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 600, in load_unet model = load_unet_state_dict(sd) ^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 573, in load_unet_state_dict model_config = model_detection.model_config_from_diffusers_unet(sd) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\comfy\model_detection.py", line 447, in model_config_from_diffusers_unet unet_config = unet_config_from_diffusers_unet(state_dict) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\rgesk\Desktop\ComfyUI_windows_portable\ComfyUI\comfy\model_detection.py", line 329, in unet_config_from_diffusers_unet match["model_channels"] = state_dict["conv_in.weight"].shape[0] ~~~~~~~~~~^^^^^^^^^^^^^^^^^^ any advice?

rafik eskander

update comfyui and load the 01 json file workflow as explained in the patreon post

Aitrepreneur

it's not a virus....altough it's certainly not the right model to dl for comfui, you need to use this one: https://huggingface.co/drbaph/FLUX.1-schnell-dev-merged

Aitrepreneur

I'll make a video about that soon, plenty of stuff to show

Aitrepreneur

VirusTotal found the link MrPittStone provided Malicious. Don't use it!!!

joh sar

Make sure to update.

joh sar

Hi Just finished downloading, finally got up and running, but on the first attempt to generate an image I get the following; Prompt outputs failed validation CheckpointLoaderSimple: - Required input is missing: ckpt_name any help would be most appericated, I'm a newby at this AI stuff, Thanks in advance,

Max Ryan

Clicking on update_comfyui.bat says: stashing current changes nothing to stash creating backup branch: backup_branch_2024-08-10_09_44_31 checking out master branch pulling latest changes Traceback (most recent call last): File "C:\progs\ComfyUI\ComfyUI_windows_portable\update\update.py", line 69, in pull(repo) File "C:\progs\ComfyUI\ComfyUI_windows_portable\update\update.py", line 19, in pull repo.checkout_tree(repo.get(remote_master_id)) File "C:\progs\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\pygit2\repository.py", line 286, in checkout_tree payload.check_error(err) File "C:\progs\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\pygit2\callbacks.py", line 99, in check_error check_error(error_code) File "C:\progs\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\pygit2\errors.py", line 66, in check_error raise GitError(message) _pygit2.GitError: 42 conflicts prevent checkout Press any key to continue . . . Clicking on update_comfyui_and_python_dependencies.bat says: stashing current changes nothing to stash creating backup branch: backup_branch_2024-08-10_09_46_35 checking out master branch pulling latest changes Traceback (most recent call last): File "C:\progs\ComfyUI\ComfyUI_windows_portable\update\update.py", line 69, in pull(repo) File "C:\progs\ComfyUI\ComfyUI_windows_portable\update\update.py", line 19, in pull repo.checkout_tree(repo.get(remote_master_id)) File "C:\progs\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\pygit2\repository.py", line 286, in checkout_tree payload.check_error(err) File "C:\progs\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\pygit2\callbacks.py", line 99, in check_error check_error(error_code) File "C:\progs\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\pygit2\errors.py", line 66, in check_error raise GitError(message) _pygit2.GitError: 42 conflicts prevent checkout - This will try to update pytorch and all python dependencies. - If you just want to update normally, close this and run update_comfyui.bat instead. - Press any key to continue . . .

cool1

It still fails for me. I tried the update bat file and the "update all" in the UI. After selecting "update all" in the comfyui manager menu: this gets shown in the command prompt: Update ComfyUI Cmd('git') failed due to: exit code(1) cmdline: git pull -v -- origin stderr: 'error: The following untracked working tree files would be overwritten by merge:' CUSTOM NODE PULL: Fail Install: pip packages FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE] All extensions are already up-to-date. After restarting the server after pressing the update, and pressing "queue prompt" it says again: "Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3']".

cool1

Can you make a Update how to to install the https://huggingface.co/sayakpaul/FLUX.1-merged model?

MrPittStone

How do you give negative prompts?

joh sar

It's about as bad at NSFW as when SD got started there are no pron checkpoints or Loras available yet.

George Thompson

the template is back here: https://runpod.io/console/deploy?template=lu30abf2pn&ref=7e9uio5y

Aitrepreneur

actually i talked to askleykza and the template is back online here: https://runpod.io/console/deploy?template=lu30abf2pn&ref=7e9uio5y

Aitrepreneur

of course, it's on the list

Aitrepreneur

yes something happened I'm trying to understand what, will let you know when I have more info

Aitrepreneur

could a multitude of reasons, first open a command prompt window and drag and drop the installer file inside the window, then press enter, this should avoid the window from closing and it will at least give us an error code we can use to troubleshoot the error, make also sure that you have disabled any antivirus/firewall from running and blocking the file/connection.

Aitrepreneur

that is weird...you should definitely have gotten a response now...have you tried using their discord server?

Aitrepreneur

there? https://www.patreon.com/file?h=109398241&i=20011741

Aitrepreneur

just update comfyui and load the 01 json file workflow as explained in the patreon post

Aitrepreneur

you don't have to redownload anything manually, all the files are hosted on my own huggingface repo anyway, just update comfyui and load the 01 json file workflow as explained in the patreon post guys

Aitrepreneur

Thank you, Nathan. I have pressed Update All, restarted GUI and it worked.

Endre Kovacs

for now yes, wait until we got nsfw finetunes :)

Aitrepreneur

you can dm me with imgur link or pasteboard.co

Aitrepreneur

whyen I try "update_comfyui_and_python_dependencies" it seems like it goes okay (but I still get the same erorr when doing "queue prompt" and "update_comfyui.bat" says "_pygit2.GitError: 42 conflicts prevent checkout".

cool1

I tried "update all" from UI option and I still get that error. I tried the .bat file too. It did come up with warnings though. Is there something that can be manually done to fix that error of " Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3']"?

cool1

Can you make a vid on Training loras for Flux would be very interesting

Terr0rClown

Its mostly sfw but can do nsfw too but not rlly good

Terr0rClown

I think you just update it, comfyui has an update file

Nathan Bourke

You have to update ComfyUI. There's an update.bat somewhere in comfyui folder

Nathan Bourke

Seems Flux is SFW only?

gillian

I tried installing the low vram version. It thinks it's installed and it says "Starting server To see the GUI go to: http://127.0.0.1:8188" and it tries to load that URL, but when it tries, though it puts "ComfyUI" in the Firefox tab name, nothing else gets shown on the page. I tried adding an exception in the antivirus in case that is stopping it but that didn't' work (no change). Do I need to add an exception in the windows 10 Pro firewall or should it s it just work? edit: I think it might be a firefox issue like before where I need to allow cookies or scripts or something as I see some html if I view the page source. edit: yes allowing cookies for that local ip address in firefox made the UI show.

cool1

I can't find it either

David

Don't expect it to make xxx stuff. But it does a really good job with anatomy likes hands and feet. Very good default realism. It can easily make chest melons and donkeys, but don't expect to find any kitty cats , swedish sausages, or x rated activities. Prompts for this are different then a lot of things on sdxl. If you want to see skin, you should mention you want skin detail, but mostly just tell it what you want in basic sentences rather than tons of single words.

Xious Divinshield

3. Choose a GPU with at least 24gb of VRAM (a cheap 3090 is great), click choose the comfyui ashleykza template, then deploy on demand The ashleykza template looks gone, it was there but cant find it anymore.

Keres666

Hey K, Long time fan, first time sub -- The FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat crashes almost immediately after Pressing A twice as instructed. At first I tried installing on my non OS drive. Then I tried installing on that same drive as Admin. Then I tried installing on my C drive, and still -- No dice. The installed crashes immediately. A fresh restart with no other Apps running isn't helping. I'm running a GeForce 4080 on Hardware about a year old. Help?

Luis Perez

I wanted to give an update. I have still received no response from Runpod about my multiple attempts to buy credits which were declined. However, I now realize that Runpod is not a good value. I can generate about 30 images per day on my own computer, and that should be enough. After running the Flux image generation for over 24 hours with my 8GB GPU, I have discovered that only 0.1 TB was written to the SSD. That is not too bad at all!

Bradford Klipan

Download the FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat... where?

Taiga

Worked with the new ae.safetensor file. Takes about 2 seconds on the schnell model with a 4090. :)

Jarryd Else

Getting same issue

Jarryd Else

what shall I do? (I've never used comfyui before)

Endre Kovacs

I'm getting an error message when using the workflow mentioned in the video: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3']

Endre Kovacs

I'm manually redownloading the ae.safetensor from the video at 4:12 during the manual install section file and putting that into the VAE folder inside models. In the video it shows a file called ae.sft in the website it has an ae.saftensors file I've dropped the ae.safetensors file into the VAE folder and in Comfy UI in the VAE node changed it to ae.safetensors After running the queue, my single 1024 image takes 14 minutes to run the schnell model Error occurred when executing VAELoader: Error while deserializing header: MetadataIncompleteBuffer File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\nodes.py", line 704, in load_vae sd = comfy.utils.load_torch_file(vae_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\comfy\utils.py", line 15, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\safetensors\torch.py", line 313, in load_file with safe_open(filename, framework="pt", device=device) as f: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Sam Shennan

I'm also getting this error: Error occurred when executing VAELoader: Error while deserializing header: MetadataIncompleteBuffer File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\nodes.py", line 704, in load_vae sd = comfy.utils.load_torch_file(vae_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\comfy\utils.py", line 15, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "A:\AI\ComfyUI\new_ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\safetensors\torch.py", line 313, in load_file with safe_open(filename, framework="pt", device=device) as f: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ I'm just using the 01 LowVRAM json workflow. So far I haven't got it to run. Also it takes so loooooooong to run on my rig 3080 10GB VRAM with 64GB RAM.

Sam Shennan

Hey, long time subscriber but generally never ask for assistance. I'm getting: Failed to validate prompt for output 9: * DualCLIPLoader 11: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] * VAELoader 10: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] Can't choose Flux in Dual Clip, and can't pick Ae.sft in the VAE loader. Please let me know what could be the issue ♥

Aaron Silehart

Just understanding the steps between installation and profit. Actually they were resolved with the v2 installer, then following the resolution of Teresa's issue.

James

Doesnt seem so, I feel like im missing something. How do I send screenshot? An outside source?

Nathan Bourke

How good is this modelk with uncencored generations Is it better

Sajjad

In the Load Fusion Model, and the DualCLIPLoader within comphyUI click and select the corresponding options, this should then correctly select the safetensor model.

James

what's the issue?

Aitrepreneur

well you need to make sure that you did, if comfyui has been succesfully updated and restarted the webui and that you correctly loaded the workflow, it should work. if it still gives you the error, then something went wrong during install, is the flux1-dev-fp8.safetensors file inside the models/unet folder? (should be a huge 11gb file)

Aitrepreneur

Anyone else struggling with steps 5.???? and 4.????

James

I believe so but I can only assume I did it correctly

Teresa Greasley

no error at all? can you send some screenshots of the webui and command prompt?

Aitrepreneur

have you updated comfyui and loaded the correct 01 json file workflow like I explained in the Patreon post?

Aitrepreneur

I tried using my old ComfyUI file and had to update it and it had the same issue, so I came here to see if this would work, same problem.

Nathan Bourke

I'm having an issue. I dont know if it's a system requirements issue or what because I have a 3060 12gb and 24gigs of ram but when I try to run it, when it starts to load the model, ComfyUI just stops with no errors.

Nathan Bourke

When I try and create an image it says "Prompt outputs failed validation UNETLoader: - Value not in list: unet_name: 'flux1-dev-fp8.safetensors' not in []" What am I doing wrong?

Teresa Greasley

Yeah, just had to install myself, that worked. Thank you so much :)

jurgis pavitols

Hi, For those whom Flux Manager auto install script is not able to download the files with an error message like "CRYPT_E_NO_REVOCATION_CHECK : This is a windows pb that sometimes requires to resolve the ssl certificates. If you are sure of the file the script tries to download (like not my responsability ;-) ), you can simply add --ssl-no-revoke at the end of every curl call in the script. Like this : curl -L -o 7z-installer.exe https://www.7-zip.org/a/7z2201-x64.exe --ssl-no-revoke Hope this helps someone. Regards and thank you again to your good work AI Entrepreneur. You rule Man !

Joseph

I see where the stuff up is. It's to do with runpod. No matter what I do to add 8888, it doesn't add. I only have 3000 and 8000. So I can never get to this terminal to do these steps.

Amon

this looks like a connection issue, make sure you don't have any antivirus or Vpn running. Otherwise you can just install git for windows yourself (it takes a few seconds) and try again

Aitrepreneur

Hi, using the FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat it gets stuck on - installing git. failed to download git installer. could you help me pls?

jurgis pavitols

I'm confused by your confusion lol, just go inside the comfyui folder, drag an drop ONE of the .sh files (like the A-Flux-fast) on the left side (the workspace) then, click on the terminal icon on the right and then copy and paste the two command lines in the patreon post for the A-flux-fast file then press enter, rewatch the video again I mean it's super easy

Aitrepreneur

I did yes. I'm broke at the "terminal" part where I copy and paste the above text. The file I open does not have the same text.

Amon

You don't have no move anything, they are supposed to be in the unet folder, you have to update comfyui and load the 01 json file workflow, I've written it in the patreon post

Aitrepreneur

Have you watched the video? I show everything there

Aitrepreneur

Probably because you ran the installer as admin so it probably installed in the system32 folder, you just need to run the installer normally and it will install in the same folder where you put the bat file

Aitrepreneur

I don't have experience with AMD gpus but I know that it works with the comfyui Zluda https://github.com/patientx/ComfyUI-Zluda

Aitrepreneur

Yes I did :-)

Mr. Illusion

Thanks for the quick response :-) FLUX-COMFYUI-MANAGER_AUTO_INSTALLV2.bat because I have not worked with ComfyUI yet. Yes, i updated "everything", to no avail

Mr. Illusion

All, Before posting a problem where you are missing clips and/or vae load the flux JSON environment and then go to Manager and hit Update All. Then restart ComfyUI. I had this problem and saw many of the error messages you are showing here until I did those steps. Thank you.

RT Solvalou

Did you update everything when performing a fresh install? I think this was answered below similarly.

RT Solvalou

Models were downloaded to the models/unet directory and where not found. I moved them to models/checkpoints

Mr. Illusion

Error occurred when executing CheckpointLoaderSimple: ERROR: Could not detect model type of: C:\Users\sd\Desktop\ComfyUI_windows_portable\ComfyUI\models\checkpoints\flux1-schnell-fp8.safetensors File "C:\Users\sd\Desktop\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\sd\Desktop\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\sd\Desktop\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\sd\Desktop\ComfyUI_windows_portable\ComfyUI\nodes.py", line 518, in load_checkpoint out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, embedding_directory=folder_paths.get_folder_paths("embeddings")) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\sd\Desktop\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 497, in load_checkpoint_guess_config raise RuntimeError("ERROR: Could not detect model type of: {}".format(ckpt_path))

Mr. Illusion

whats this terminal business? can i just git the runpod file?

Amon

Where is supposed to install to? drive folder location? its installed, but i have no idea where, its not the same folder location I ran the installer from

Frank B

Just want to ask, can it work with an amd gpu ?

Sayed

I have the same issue, i am running it on a different drive that C:

Jacob Guité

Should have read the post properly! I apologize

Keiron Gulrajani

read the first line in the patreon post, you need to update comfyui and load the correct workflow

Aitrepreneur

what's the issue?

Aitrepreneur

I'll check it out

Aitrepreneur

Hey I ran the 1 click installer on my machine, and loaded the json into comfy ui but i get the following error when i queue the prompt: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] EDIT: I seem to have sorted it by running the update_comfyui_and_python_dependencies.bat which updated everything that was out of date

Keiron Gulrajani

I cant even seem to be able to run the comfyui installer correctly...I'm starting to think my system is borked

Twisted Tiger

hey aitrepreneur! i recently discovered dough from banodoco and i thought maybe you can make a video about it? there are also comfyui nodes called motion director and steering motion which i believe are used in dough... im having trouble setting it up though. Would be sick!

MarcAndreMueller

I have EXACTLY 32gb of RAM, and yes....schnell does indeed run faster, rendering in about 1 minute, 15 seconds on average. Even after making sure I updated my video drivers, the regular fp8 model is still averaging about 5 minutes. Oh, well....it's all fine. I mean, as you said, even schnell looks pretty spectacular! Thanks for the help.

Ken L. Black

send me in dm a screenshot of the webui and check if the models have been correctly installed into the correct folders also which installer file did you use on runpod and which workflow?

Aitrepreneur

I keep getting this eror when trying the Scnell model in Runpod, even though ComfyUI is updated. I tried all the workflows and nothing seems to work :( Error occurred when executing UNETLoader: Error while deserializing header: HeaderTooLarge File "/workspace/ComfyUI/execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/nodes.py", line 836, in load_unet model = comfy.sd.load_unet(unet_path, dtype=dtype) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/comfy/sd.py", line 626, in load_unet sd = comfy.utils.load_torch_file(unet_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/comfy/utils.py", line 15, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/venv/lib/python3.11/site-packages/safetensors/torch.py", line 313, in load_file with safe_open(filename, framework="pt", device=device) as f: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Петр Баринов

I was using the wrong json. Derp. Thanks for your help!

Duncan

Send me a dm, use imgur or pasteboard.co

Aitrepreneur

have you loaded the correct json worflow file?

Aitrepreneur

how much ram you got? If you have less than 32gb of ram then yes 5mins is the speed you get with your current config unfortunately, you can try using the schnell model with 4 steps and generating at lower res or rent a gpu on runpod for a few cents as well

Aitrepreneur

hi. I have downloaded everything from the download link, but I DO NOT END UP WITH THE SAME AS YOU. I wanted to send a screenshot, but there doesn't seem to be an option. Am I doing sometrhing wrong?

mike atherfold

Hi, getting error when running it. Error occurred when executing SamplerCustomAdvanced: mat1 and mat2 shapes cannot be multiplied (1x1280 and 768x3072). Have "Updated All" as advised error still occurs. Help appreciated thnx.

Duncan

Thanks for the quick reply. I THOUGHT so! I'm using the one you showed in the video with the unet name "flux1-dev-fp8.safetensors" with weight_dtype of "fp8_e4m3fn" for Load Diffusion Model. The DualCLIPLoader has clip_name1 of "clip_l.safetensors" and clip_name2 of "t5xxl_fp16.safetensors," with type "flux." Everything else is same. This is for the default width/height of 1024x1024, so I didn't change that, either. The only thing I really did was type in my own prompt and set noise_seed to 0 (zero), to ensure it was random. I appreciate any insights you can give me. I mean, yeah....I can *live* with 5 minutes, but I sure thought that 12GB would give me a BIT faster speed than that!

Ken L. Black

Yeah, I edited the comment to make that more clear. Thanks for all the cool stuff!

Ameshin

this only works if you have 7zip correctly installed though, if it is then yes

Aitrepreneur

definitely a bit too slow for 12gb vram, are you using the correct 01 workflow json file? have you also tried the schnell model with 4 steps as well?

Aitrepreneur

Make sure that you use the normal run_nvidia_gpu.bat file and that you have selected the 01 workflow and that you have selected the fp8 version of the t5xxx encoder. Make also sure just like in the video that you have disabled sysmem in the nvidia control panel

Aitrepreneur

I suggest you contact runpod customer support they are really nice and super helful they will help if you need to add credits, will be much faster than using your 8gb vram GPU

Aitrepreneur

Use this workflow for negative prompts: https://we.tl/t-NNuBLQrwen It works....ok but unfortunately because this is a hack, this basically made the generation twice as slow First update comfyui, otherwise it won't work. Load the 03 workflow json file, then go to manager, click on "Install missing Nodes manager", there you will see 2 entries: Dynamic Thresholding Extra Models for ComfyUI Click on the checkbox to select the 2 entries then click install then click restart to restart the webui. Then you can write a prompt, add negative prompts

Aitrepreneur

To anyone having issues with the comfyUI installer and has 7zip correctly installed already I just deleted the part in the batch (bat file) where it checks and it worked. -- This part: REM Check if 7-Zip is installed and get its path for %%I in (7z.exe) do set "SEVEN_ZIP_PATH=%%~$PATH:I" if not defined SEVEN_ZIP_PATH ( if exist "%ProgramFiles%\7-Zip\7z.exe" ( set "SEVEN_ZIP_PATH=%ProgramFiles%\7-Zip\7z.exe" ) else if exist "%ProgramFiles(x86)%\7-Zip\7z.exe" ( set "SEVEN_ZIP_PATH=%ProgramFiles(x86)%\7-Zip\7z.exe" ) else ( echo 7-Zip is not installed. Downloading and installing... curl -L -o 7z-installer.exe https://www.7-zip.org/a/7z2201-x64.exe 7z-installer.exe /S set "SEVEN_ZIP_PATH=%ProgramFiles%\7-Zip\7z.exe" if not exist "%SEVEN_ZIP_PATH%" ( echo Installation of 7-Zip failed. Please install it manually and try again. exit /b 1 ) del 7z-installer.exe ) )

Ameshin

So everything seems to be running fine (and yes, I'm clicking on run_nvidia_gpu.bat to run ComfyUI), but when I download your JSON workflow and run it, renders are taking over 5 minutes. I'm on an NVIDA GeForce RTX 3060 12GB of VRAM. Would this be normal, do you think? If not, what am I doing wrong?

Ken L. Black

Yeah that's good, but to make it even better... insert it as a step in the instructions. (I'm speaking from personal experience.. give it a shot, it works)

Alan Smile

I get message saying: loading in lowvram mode. Is this normal? I'm on a 4090 and it feels really slow (like +40s)

Silly Pear

Update. I believe I got everything working. However, Runpod has a flakey payment processor and they won't let me buy any more credits, therefore RunPod is no longer an option. Here is the info on my existing options to use Flux with the highest quality models: (CPU Only) Core I5 8th generation, 64GB RAM: 58GB Ram Used By ComfyUI, Image size 1168x1536, 11 minutes per iteration, 6 hours and 25 minutes per 35 step image. Gtx 1070-ti (8gb), 16Gb RAM, 120GB SSD Paging file, Image size 1168x1536, 80 seconds per iteration, 47 minutes per 35 step image (Heavy SSD Usage For Paging!). These are my only options, but at least they work and are better than nothing. I realize that my system with the GPU makes heavy use of the SSD for virtual memory paging and will wear out the SSD prematurely if used for a long period of time. Therefore, I will use my GPU system for SDXL generations, and my CPU system could be used for Flux generations.

Bradford Klipan

What are some good samplers and schedulers to use? I get blurry images when I use the ones I normally try on Fooocus. Also, how to add negative to ComfyUI?

Miguel NoName

yup then everything is working as intended with your GPU. If you wanna be 100% sure then download a simple SD 1.5 model and use the default workflow and see how fast it takes for you to generate an 512x512 image, it should only take like a few seconds, a CPU wouldn't be able to generate an image this fast

Aitrepreneur

the default image of the lion takes roughly 5-6 minutes total. After switching to the Schnell version and setting the number of steps to 4 as suggested in the video, roughly 45-60 seconds

Some Guy Named Rob

might be because you have 8gb so it's offloading the majority of the task to ram, maybe it's not displaying that properly as well, how long does it take to generate an image?

Aitrepreneur

Task manager is showing 0% utilization of the GPU during the image generation, and my CPU pins to 100%. The device from the bat file launch is showing my GPU correctly, but Comfy UI isn't making use of it at all from what I'm seeing. I have a screen capture of the generation and my RAM and CPU are being well used, my GPU is doing nothing. Bat file output: C:\AI\ComfyUI_windows_portable>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --lowvram [START] Security scan [DONE] Security scan ## ComfyUI-Manager: installing dependencies done. ** ComfyUI startup time: 2024-08-05 11:59:17.574627 ** Platform: Windows ** Python version: 3.11.8 (tags/v3.11.8:db85d51, Feb 6 2024, 22:03:32) [MSC v.1937 64 bit (AMD64)] ** Python executable: C:\AI\ComfyUI_windows_portable\python_embeded\python.exe ** ComfyUI Path: C:\AI\ComfyUI_windows_portable\ComfyUI ** Log path: C:\AI\ComfyUI_windows_portable\comfyui.log Prestartup times for custom nodes: 1.1 seconds: C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager Total VRAM 8192 MB, total RAM 16064 MB pytorch version: 2.3.1+cu121 Set vram state to: LOW_VRAM Device: cuda:0 NVIDIA GeForce RTX 3070 Laptop GPU : cudaMallocAsync Using pytorch cross attention [Prompt Server] web root: C:\AI\ComfyUI_windows_portable\ComfyUI\web ### Loading: ComfyUI-Manager (V2.48.5) ### ComfyUI Revision: 2469 [a178e259] | Released on '2024-08-05' Import times for custom nodes: 0.0 seconds: C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py 0.3 seconds: C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager Starting server To see the GUI go to: http://127.0.0.1:8188 [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/alter-list.json [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json FETCH DATA from: C:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json [DONE] [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/custom-node-list.json got prompt model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16 model_type FLOW clip missing: ['text_projection.weight'] Requested to load FluxClipModel_ Loading 1 new model Requested to load Flux Loading 1 new model loading in lowvram mode 5414.064

Some Guy Named Rob

What makes you say that? What does it say on the "device:" line when starting the bat file? If your GPU is listed there then that means it's using your GPU normally, also as I said with 8GB of vram it's gonna take a while to generate, Flux is a big ass model...

Aitrepreneur

I was able to get everything working using the bat files and included tutorial, but my images are rendering from my CPU instead of my 3070 (8GB) GPU. Any suggestions would be most helpful to speed things up.

Some Guy Named Rob

Haha well not if you just click "edit with notepad" it's not an exe file, you can literally just look at the code inside and see for yourself without any risk

Aitrepreneur

Yeah but bats can carry rabies. Opening it up to look inside seems like the opposite of safe. 🦇

Carl Brahms

fixed it thank you!

Matt

make sure that for your setup, you are using the t5xxl_fp8_e4m3fn.safetensors instead of the fp16, it will speed up the generation process

Aitrepreneur

I have an interesting issue, anytime I write a new prompt the generation takes 1+ minute but if I leave the prompt after that it takes 14 seconds. Running off a RTX 4090

Matt

try opening a new cmd prompt and drag and dropping the bat file inside and pressing enter, this might avoid the cmd window from automatically closing and it will at least give us an error or something to work with. A simple crash with no other info is not really helpful to find the cause

Aitrepreneur

like i said, its not generate a error, just stop the cmd.exe. Its appear only with schnell, with dev its workink, a idea ?

Sioko

are you sure you are using the correct workflow? It's looking for the flux1-dev model meaning that you loaded the 02 Flux json file instead of the 01

Aitrepreneur

It's a bat file, you can look inside if you don't trust me lol

Aitrepreneur

I know nothing about blender but yes, I suggest however that you look for SheepIt, it's supposed to be a free render farm that you can use instead

Aitrepreneur

can you give me some screenshots or something? Without any error log is difficult to troubleshoot

Aitrepreneur

that's why I put the first line in the post to update comfyui and load a workflow, so that people knows that a simple solution exist for the error they are getting

Aitrepreneur

UI updated. Workflow loaded. Now I get an error "Prompt outputs failed validation UNETLoader: - Value not in list: unet_name: 'flux1-dev.sft' not in ['flux1-dev-fp8.safetensors', 'flux1-schnell-fp8.safetensors']" What do?

Billy Brujo

Has someone vetted this code?

Aryan Saxena

Aitrepreneur, you have lots of experience in runpod. and I have this off topic question that noone has answer to, but do you think renting a powerful card help me in Blender rendering? Thanks again!

Mohamed Owda

This has indeed fixed the issue! Thank you very much :)

Pumpalooza

but when I copied the model to the folder I could run the flux model (Im not choosing the 1.5 SD Im choosing the flux in ComfUi) weird :D so maybe its a coincidence that it started working ?

Peter Rossa

Hey, the step in Comfyui crash on VAE Decode without error, juste Loading 1 new model and pause... its a new install auto, all files is on models repertory, and i try with run_nvidia_gpu-LOWVRAM.bat, schnell and 4 steps Have a card 4070 super with 12Go Vram and 32 go ram

Sioko

The main issue aside, I think the reason for all the comments is that it's not totally obvious to click "Load Replies" for those new to Patreon, so at first glance it looks like no solution is available.

John Cabrer

again same advice as the others, just update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column. Then load a workflow file (like the 01 json file in this post)

Aitrepreneur

Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] Where to place flux folder in comfyui? Confused

Toime

again same advice, update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column. Then load a workflow file (like the 01 json file in this post)

Aitrepreneur

Thank you very much! That did it! :)

Len

Hi there, Boss! I do get the exact same error. I can say that I did the automatic installation and made sure all the files of the models were placed correctly, as I was following the tutorial. Any suggestion? Thanks a lot for your time :) I appreciate the help.

Pumpalooza

okay got it to work! Thank you so much!

Chris D

Is this a fresh install? Have you restarted the webui? Are the models correctly downloaded into the correct folders?

Aitrepreneur

I loaded the json, let me try and gen image its taking a minute.

Chris D

Fully updated ComfyUI Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3']

Trey G

Ok then now update comfyuu by clicking on the manager button and then clicking update all, then restart the ui, load a json workflow file present in this post then try generating an image

Aitrepreneur

Comfy starts up. But no Flux

Chris D

what happens when you manually run the normal "run_nvidia_gpu.bat" file present in the ComfyUI_windows_portable folder?

Aitrepreneur

Thank you. I resolved the issue by downloading and running the powershell version of your script. There were still some problems, but I was able to adapt solutions from similar issues posted here.

Truman Brown

When i try regular install I get error: 'Entry' is not recognized as an internal or external command, When I try v2 I get error: '"run_nvidia_gpu-LOWVRAM.bat"' is not recognized as an internal or external command, operable program or batch file.

Chris D

could a multitude of reasons, first open a command prompt window and drag and drop the installer file inside the window, then press enter, this should avoid the window from closing and it will at least give us an error code we can use to troubleshoot the error, make also sure that you have disabled any antivirus/firewall from running and blocking the file/connection.

Aitrepreneur

first update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column and then load a json file workflow provided in the post

Aitrepreneur

I get the same error, just installed it using the script.

Mats Danielsen

install is not working. v2 either. just closes out.

Chris D

The workflow are the json files present in the post: 01-FLUX-FAST-LOWVRAM.json 01-FLUX-IMG2IMG-FAST-LOWVRAM.json 02-FLUX.json 02-FLUX-IMG2IMG.json

Aitrepreneur

I did? What was the issue?

Aitrepreneur

Also can you provide a "working" workflow?

Einziger

Not working for me. i did a new install using the script. Try to install a Workflow (not included????) and finally got this error: Error occurred when executing UNETLoader: 'conv_in.weight' File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data uis = [] File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list nodes.before_node_execution() ^^^^^^^^^^^^^^^^^^^ File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\nodes.py", line 829, in load_unet dtype = None ^^^^^ File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 600, in load_unet diffusers_keys = comfy.utils.unet_to_diffusers(model_config.unet_config) ^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 573, in load_unet_state_dict File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\comfy\model_detection.py", line 447, in model_config_from_diffusers_unet 'num_classes': 'sequential', 'adm_in_channels': 2816, 'dtype': dtype, 'in_channels': 4, 'model_channels': 320, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\Ixoneo\Downloads\ComfyUI_windows_portable\ComfyUI\comfy\model_detection.py", line 329, in unet_config_from_diffusers_unet t_out = [] Queue size: 0 Extra options

Einziger

Please make a new tutorial on how to use Flux with runpod. Clearly there are errors in the instructions you gave us. Please also explain how to save the installation to network storage. I was able to generate only one image on runpod and then afterwards I only got errors.

Bradford Klipan

I’ll try to re-download the model

Robb

ok were all the models downloaded correctly? This error usually means that there might be a corrupted model here...

Aitrepreneur

It’s a new install using the full install script.

Robb

Is that a new or old install?

Aitrepreneur

Thanks for the info.

Bradford Klipan

Did that. Still not working.

Robb

have you seen the video? I show everything there

Aitrepreneur

but I did, .sh file is for runpod and I tell you to go inside the comfyui folder and put one the 3 .sh files there

Aitrepreneur

what workflow have you loaded? you need to load the 01 json file that is in this post and make sure that you updated comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column.

Aitrepreneur

Make sure to update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column.

Aitrepreneur

what? No, it's the opposite. That's a bat advice, this means you are not using the correct model and workflow...load the 01 json file workflow no need for the 1.5 model...you are not using SD here you are using Flux...

Aitrepreneur

yeah ok I found the issue, it's fixed now in the new V2 installer, the problem was indeed the lowvram bat file

Aitrepreneur

Bradford Klipan

I've looked into your .sh files and noticed that they are expected to be run from ComfyUI folder, but you haven't mentioned it. I think it would be nice to update your tutorial a bit.

Bob Bobenko

this worked for me. thank you!

Kyle

My LOWVRAM bat file contains the same text

Trey G

No errors during install and I did the update all in manager because at first it didn't accept flux only sd, sdxl and sd3. So the update allowed flux and the appropriate json file that you recommended for 12gb cards and I use the same prompt the lion one then I hit the generate and o get that message in a box in red font

james

I ran FLUX_AUTO_MODELS_INSTALL.bat in an existing comfyui. When I load the workspace, ae.sft does not show up in the LoadVAE module, even though I can see it is in the vae folder. Same with flux1-dev.sft

Gary

I'm getting this error and I have a 3090 Error occurred when executing UNETLoader: Error while deserializing header: MetadataIncompleteBuffer File "E:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\ComfyUI_windows_portable\ComfyUI\nodes.py", line 836, in load_unet model = comfy.sd.load_unet(unet_path, dtype=dtype) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 626, in load_unet sd = comfy.utils.load_torch_file(unet_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\ComfyUI_windows_portable\ComfyUI\comfy\utils.py", line 15, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\safetensors\torch.py", line 311, in load_file with safe_open(filename, framework="pt", device=device) as f: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Robb

Well it then terminated the terminal. I couldn't generate images

Jim Lim

that error shouldn't impact the generation at all, I got the error as well on my side

Aitrepreneur

I'm getting this error: got prompt model_type FLUX model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16 clip missing: ['text_projection.weight'] I'm using a Macbook and had manually installed the files from huggingface. Thoughts?

Jim Lim

Sure - it is here https://pastebin.com/Li7mJRx0

TAGILUX

its working, I read in the our discord that I have to download sd 1.5 ckpt model and put it into checkpoint model, and its working

Peter Rossa

Can you give me the full error log?

Aitrepreneur

Thanks - but this is on install and the Flux Model is not downloaded as a result of the installer failing at that point

TAGILUX

Dl the lowvram bat file from here if you need it: https://we.tl/t-KuusC74aiq then put it in the ComfyUI_windows_portable folder

Aitrepreneur

here's the bat file if you need it: https://we.tl/t-KuusC74aiq

Aitrepreneur

Have you updated comfyui yet? (click on the 'Manager' button on the right side, then 'Update All' from the center column)

Aitrepreneur

Make sure to update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column.

Aitrepreneur

here's the bat file if you need it: https://we.tl/t-KuusC74aiq

Aitrepreneur

That lowvram file error is super weird, if you need the bat file here it is: https://we.tl/t-KuusC74aiq And Billy as I said you are using the wrong workflow and the wrong model, load the correct Json worflow (01 for example) present in this post

Aitrepreneur

You need to load a workflow (like the 01 json file in this post) and also update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column. The error you got is by using the default workflow.

Aitrepreneur

Update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column.

Aitrepreneur

you should update comfyui by clicking on the 'Manager' button on the right side, then 'Update All' from the center column. Much easier and faster this way

Aitrepreneur

you need to first, load a workflow (like the 01 json file in this post) and also update comfyuii by clicking on the 'Manager' button on the right side, then 'Update All' from the center column. The error you got is by using the default workflow, you downloaded the stable diffusion 1.5 model, that you don't need to use here

Aitrepreneur

Same. Low rez doesnt work. Regular loads fine... but ... tried some prompts and its not very good?? Do I need a different model or a different checkpoint? It seems like a downgrade from the stablediffusion I was using?????

Billy Brujo

here: https://we.tl/t-KuusC74aiq but with 8gb vram that's gonna take time anyway, that model is huge, definitely try the schnell version as well with 4 steps for faster generation

Aitrepreneur

What error does it give you when installing 7zip? Try first uninstalling 7zip from your pc then relaunching the installer, make sure you don't have any vpn or antivirus running that could block the connection as well

Aitrepreneur

got any errors during install? when exactly does that happen?

Aitrepreneur

yes it's running locally, nothing is send out anywhere

Aitrepreneur

can you send screenshots? That's definitely no normal

Aitrepreneur

well what is the error you got?

Aitrepreneur

Try installing git manually if the auto doesn't work, then relaunch the installer

Aitrepreneur

Anyone get this when running the FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat ComfyUI and FLUX installed. Running ComfyUI... 'Entry' is not recognized as an internal or external command, operable program or batch file.

TAGILUX

Same.

rfromp

run_nvidia_gpu - LOWVRAM.bat doesn't work. I see a very brief flash of the command window before it closes. When I open run_nvidia_gpu - LOWVRAM.bat all it says in that file is: "Entry not found"> Something is missing in that .bat file.

rfromp

same error for me, (I created the sam e worklfow compared it to Your youtube vid)

Peter Rossa

RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x1280 and 768x3072)

Peter Rossa

I folks! I installed the FLUX_AUTO_MODELS_INSTALL.bat cause I've already installed comfyui. However there is an issue when queque: Error occurred when executing UNETLoader: 'conv_in.weight' How can i solve this?

Andrea Tamburrini

ok man i have the fix for you, open ComfyUi, open Tab "Manager", in the middle column click "Update ComfyUI", restart and you are ready to go

WildKeeper

same problem for me, for the VAE and the Unet Model i think is becouse they are .sft instead of .safetensors, i changed them manually and now you can select them, but i still don't know why the "flux" type is not showing in the DualCLIPLoader

WildKeeper

Open the folder in CMD, then run the file from that instance and it does load the interface. However, it doesn't seem to load the model in this interface (looking for a checkpoint file that doesn't exist instead). Probably because the LOWVRAM.bat file is not correct in the distro.

Dustin

For everyone having issues with LOWVRAM.bat file, open it in notepad, post this and save then run: .\python_embeded\python.exe -s ComfyUI\main.py --lowvram --windows-standalone-build pause If you need the workflow for Flux, it can be downloaded here. Just save it in notepad as workflow.json then drag and drop it into the comfy interface. You may need to point to your safetensor model files. https://drive.google.com/file/d/1v8xd97Qz0gB46xH7wFLsSgePPNMTkF1c/view

Dustin

Well there's a "Git-2.41.0.3-64-bit" on the folder now. Do I have to install it?

Netierrez

Even this one goes to this: https://ibb.co/rtvLVmQ and Crashes. (I already run the one in this post, not sure I have to uninstall it before installing this new one. But I also dont know how to uninstall it LoL)

Netierrez

Same Here =(

Netierrez

Following the procedure "LOCAL INSTALL If NO comfyUI is installed", I am stuck at step four. The file run_nvidia_gpu-LOWVRAM.bat is missing from the ComfyUI_windows_portable folder.

PN Houle

same, LOWVRAM has only one line "Entry not found" @AITREPENEUR - please help :)

JM

To those having trouble with install, make sure you run the .sh install in /ComfyUI in terminal, mine did not go there automatically, and was in /workspace. It couldnt find the models and unet folders because they were in a different directory.

Jayden Kirkham

Correction, LOWVRAM only gives me a flash and closes, but the regular nvidia_gpu works for me. If I try to edit "run_nvidia - LOWVRAM" there is only: Entry not found

Productions Diratia

Same here

Productions Diratia

"Wouldn't even load. Small Window Just flashes and closes before anything happens". Same thing with me.

totally_not_a_weirdo

Wouldn't even load. Small Window Just flashes and closes before anything happens. Checked readme file. It says make sure you have a checkpoint in the checkpoint folder. Says load the SD 1.5 one. Sure enough Checkpoint folder empty. Load the SD 1.5 in there. Still doesn't work. Commandprompty-type window pops up for halfsecond and closes. What have I done wrong?

Billy Brujo

I renamed ‘ae.sft’ - “ae.safetensors” Make sure to update ComfyUI

Lord Banzai

I renamed ‘ae.sft’ - “ae.safetensors” Make sure to update ComfyUI

Lord Banzai

I'm getting this error Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] How do I fix this?

XxForTheKillxX

Managed to make it work by: ``` cd ComfyUI git pull ``` Then exec your proper .bat file

Romário Lima

Thank you so very much! I reinstalled using the batch file twice both fresh installs. I didn't even think to update all in the manager. I thought there was something wrong with my system. Thank you!

Gaming For The Masses

Well, it did eventually run: "Prompt executed in 648.04 seconds" LOL

Scott Groll

got prompt model_type FLUX model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16

Scott Groll

Everything loads as expected. However, when I press "Queue Promp" all I get is:

Scott Groll

I never used ComfyUI before. After ran FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat and launched run_nvidia_gpu.bat my ComfyUI window is missing Load Diffusion Model and DualCLIPLOader completely :( Also I was missing checkpoints so I downloaded v1-5-pruned-emaonly.ckpt and placed into checkpoints folder. Image generation looks hideous and I can't change models. Please help.

Len

got it working eventually by clicking queue front and then it genned both.

Connor Hair

This worked! Thanks a lot!

Philip Geiblinger

also says this. clip missing: ['text_projection.weight']

Connor Hair

Updating fixed it, thanks!

closeded

got that working by using the V2 installer but now It just hangs after queueing the prompt. last thing in the cmd is model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16

Connor Hair

I'm not seeing the LowVRAM.bat, can it be added to the post please? I got it working by downloading comfyui myself, then dropping the auto installer into the comfyui folder and running the .bat there, it's just extremely slow on my 8gb VRAM gpu. 5 minutes for 512x512...

NathanCA2014

thanks dude. it helped me as well

Dark Dude Maurycy

yoo. go to manager and update all. this fixed it for me

Dark Dude Maurycy

same here. for both 01s and 02s

Dark Dude Maurycy

when i run the FLUX-COMFYUI-MANAGER_AUTO_INSTALL and choose A for each option it just closes without installing.

Connor Hair

Hi. I've tried both the original and the V2 version of the ComfyUI installer bat file and both continue to fail citing that 7-Zip is not installed. When it attempts to download and install 7-Zip - it fails. 7-Zip was already installed on my system and I've also tried manually re-installing it again (successfully), but the bat file continues to say that 7-zip is not installed. Please advise. Thanks!

Truman Brown

Runt he 7z installer then rerun the bat

Trey G

Error occurred when executing UNETLoader: Error while deserializing header: MetadataIncompleteBuffer File "E:\coumfy ui and flux\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\coumfy ui and flux\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\coumfy ui and flux\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\coumfy ui and flux\ComfyUI_windows_portable\ComfyUI\nodes.py", line 836, in load_unet model = comfy.sd.load_unet(unet_path, dtype=dtype) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\coumfy ui and flux\ComfyUI_windows_portable\ComfyUI\comfy\sd.py", line 626, in load_unet sd = comfy.utils.load_torch_file(unet_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\coumfy ui and flux\ComfyUI_windows_portable\ComfyUI\comfy\utils.py", line 15, in load_torch_file sd = safetensors.torch.load_file(ckpt, device=device.type) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\coumfy ui and flux\ComfyUI_windows_portable\python_embeded\Lib\site-packages\safetensors\torch.py", line 311, in load_file with safe_open(filename, framework="pt", device=device) as f: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

james

I did this to fix it too. Then one more bug came I had to close and relaunch then it works fine for multiple generations.

Jason Graham

SD though Automatic 1111 does not reach out to the internet by default. I won't share my generations or prompts to a public server. It works without internet soon. If ComfyUI the same?

Jason Graham

I think this worked for me too. It's now loading stuff when I click Queue prompt etc. Thanks.

Jason Graham

When I load "01-FLUX-FAST-LOWVRAM.json" """ Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] """

closeded

I loaded 01-FLUX-FAST-LOWVRAM even though I have a 4080. All four workflows I found of your give similar errors. The "01-FLUX-FAST-LOWVRAM" gives. Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] First time ComfyUI user so I don't know what I am doing wrong. Maybe something failed download.

Jason Graham

Same issue here? Is there like a VAE and clip loader and that we need to download manually or something? That the one click failed to load? First time ComfyUI user and first time Flux user here.

Jason Graham

Everyone getting that error just goto Manager and Update All

Lloyd Braun

Unfortunately the same crash... I turnt off as much as I could for any firewalls and antivirus on windows, unless you have something specific I should check?

S G

run_nvidia_gpu - LOWVRAM.bat contains "Entry not found" text instead of a script to run with low vram. Is it a mistake?

Ihor Aleinikov

I can't get it to work. I have been trying the different workflows that were posted. I have an RTX 4080. For some reason, run_nvidia_gpu - LOWVRAM also does nothing.

Diggy Dre

im getting the top two errors as well

Jordan

That's what I was using as the video stated.

Kyle

Yes

Aitrepreneur

to clarify do you mean 02-flux.json?

Kyle

have you loaded the correct workflow?

Aitrepreneur

that means you chose to dl the normal model, then you need to load the 02 workflow

Aitrepreneur

because you loaded the wrong workflow, load the 01 instead

Aitrepreneur

Have you tried this one: https://we.tl/t-Uh2DHzrscf Make also sure that you don't have any antivirus/ vpn running as well

Aitrepreneur

Models said they are downloaded. I have a 23gb file called flux1-dev.sft and 11.6 gb flux1-schnell-fp8.safetensors in the unet folder

Kyle

I'm getting this error: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] UNETLoader: - Value not in list: unet_name: 'flux1-dev.sft' not in ['flux1-schnell-fp8.safetensors']

Kyle

RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x1280 and 768x3072)

ClearanceClarence

Ran installer. Chose A, A. Ran run_nvidia_gpu.bat In comfy UI, Load -> 02-FLUX.json Queue Prompt """ Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] UNETLoader: - Value not in list: unet_name: 'flux1-dev.sft' not in ['flux1-dev-fp8.safetensors', 'flux1-schnell-fp8.safetensors'] """ Did I miss a step?

closeded

yep same here

Jordan

Same here :(

NathanCA2014

1. Download the FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat 2. Run the bat file and select A for both choices 3. CRASH... and now I have a 7z-Installer in the same folder. I use 7z a lot so IDK why i'd need to install it again, also I have tried all the other .bat files you've posted hoping ComfyUI would install but same thing.

S G

Downloaded: FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat Prompt outputs failed validation DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] I also had a problem with the model in Unet and the vae model, they are saved as .sft files, and when selecting them they did not appear in the list, I changed the name by hand writing .safetensors , instead of .sft, now they appear

WildKeeper

For anyone getting the "Value not in List" errors, just update comfyUI. That fixed mine. Click 'Manager' on the right side, then 'Update All' from the center column.

Dillon

updating comfy worked

ty110072

Use the json files provided in the post

Aitrepreneur

I dont provide installers for osx since i don't have any apple product at home so I cant make sure that it works

Aitrepreneur

Thank you !, this one work for me

Anders

i am getting the same error: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] Queue size: 0 Extra options

ty110072

Make sure that comfyui is up to date if you had it downloaded already

Aitrepreneur

Same here. I tried loading all the JSON files provided and get different errors each time

remy wilders

After I installed Comfy UI and everything with the installer, if I run comfy what file should I load and where is it located?

bum

Hi, is there installer for OSX?

David Rynes

SOLVED. comfyui needed to be updated to work. i needed to click update all and update comfyui in the settings, then restarted.

Natale Fin

Having the same problem, you said it at 2:40 in the video, what am I supposed to load, sorry I didn't quite get it.

bum

The following folder is empty: ComfyUI_windows_portable\ComfyUI\models\checkpoints

remy wilders

I got a bunch of errors in comfyui when I tried to queue a prompt.

Marcus Sousa

I must have missed something. Where do I select the workflow ? I simply clicked on default workflow

remy wilders

are you sure you used the right workflow? This is looking for the SD 1.5 model that shouldn't be used here...

Aitrepreneur

as I said in the video, this doesn't support any training for now

Aitrepreneur

does this support lora's?

John

I'm able to set every thing up and I load the default workflow but then I get the error : Prompt outputs failed validation CheckpointLoaderSimple: - Value not in list: ckpt_name: 'v1-5-pruned-emaonly.ckpt' not in []

remy wilders

well usually if there was not previous install it should install python and git automatically, but I might remove it from the installer since it can conflict with a previous install, but I'm here to help anyway just in case, so have fun

Aitrepreneur

ahh, now I get it. I need to install Python and Git manually and then have to run the batch file again. now it is installing comfy UI. I am installing it on a clean Windows 10 install... Thank you Aitrepreneur for the fast reply!

JM

was the installation successful? have all the models downloaded correctly? You can send me some screenshots as well

Aitrepreneur

followed the instructions with the 1 click installer now get this when loading your wokrflow: Prompt outputs failed validation VAELoader: - Value not in list: vae_name: 'ae.sft' not in ['taesd', 'taesdxl', 'taesd3'] DualCLIPLoader: - Value not in list: type: 'flux' not in ['sdxl', 'sd3'] UNETLoader: - Value not in list: unet_name: 'flux1-dev.sft' not in ['flux1-dev-fp8.safetensors', 'flux1-schnell-fp8.safetensors']

Natale Fin

ok...then try this powershell version: https://we.tl/t-Uh2DHzrscf

Aitrepreneur

well, now it says installing git..., downloading, then stops

JM

it's not that hard really, just upload a workflow and everything is already setup for you, it's harder to use than auto1111 but for basic generation it is as easy

Aitrepreneur

try this version: https://we.tl/t-Uh2DHzrscf

Aitrepreneur

try this one: https://we.tl/t-DAywKzWRb0

Aitrepreneur

try this one: https://we.tl/t-DAywKzWRb0

Aitrepreneur

try this one: https://we.tl/t-DAywKzWRb0

Aitrepreneur

LOCAL INSTALL If NO comfyUI is installed : 1. Download the FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat 2. Run the bat file and select A for both choices (my recommendation) CRASH - What next?

Marcus Sousa

hmm. tried the FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat and after entering the A option twice the bat file just exits... B:\AI\FLUX>B:\AI\FLUX\FLUX-COMFYUI-MANAGER_AUTO_INSTALL.bat Choose your preferred installation: A) Fast-Lowvram install B) Unoptimized normal model Enter your choice (A or B) and press Enter: A Do you want to download the FLUX SCHNELL Model? A) YES B) NO Enter your choice (A or B) and press Enter: A B:\AI\FLUX>

BS TUEJJ

Hi, the comfyui Auto installer stops after downloading python. Any ideas?

JM

Flux looks good and I tried the demo online (sometimes text wasn't added exactly as it said in the prompt though), though I read it can't do Loras. Not sure if it will in future but it's a shame if it can't but SDXL can. ComfyUI seems quite a bit more complex than automatic1111 too (and there's been some issues with some of the nodes like some people added bad nodes I read). Hopefully it won't be too complex and will be okay.

cool1

"The System Cannot Find the Path Specified." is what I got for all the files(VAE, CLIP...) also it says "'curl' is not recognized as an internal or exterenal command"

Onur Tumer