🤝｜tech-support

[remote cosmos]

Yeah this is my first time working with ControlNet so Im not even sure where to start troubleshooting

[ornate elk]

Try use the 700mb openpose model from the bottom here:
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[reef smelt]

My image quality is pretty grainy, why is that?

I'm using dreamshaperXL_v21TurboDPMSDE

With auto1111 forge

my prompt is very simple, should sdxl models have vae's?

[remote cosmos]

Hrm that one doesn't seem to do much

it cooks the colors a little less though which is nice lol

@ornate elk @ruby yoke Ayyy I think I figured it out

I have no idea what a pre-processor does, outside of help generate that preview

But I found that turning the preprocessor OFF and using the "full thibauld" model works!

[gloomy bolt]

question, recently my images have been generating with like a light rainbow mosaic or dots in the background

any ideas as to what the issue might be?

[remote cosmos]

What happens if you turn off the upscaler?

[gloomy bolt]

the hires.fix?

[remote cosmos]

Yeah, what do your native resolution generations look like

Ive had upscalers mess my images up a lot before

[ruby yoke]

you know what you should do now? create a tutorial for this - and that way others with the same problem can find a good, fast answer

[gloomy bolt]

it just makes the rainbow mosaic even stronger and worse

[remote cosmos]

Really? Weird, most of my "mosaic" or "tile" issues were from the upscaler

[ruby yoke]

change your sampler

[gloomy bolt]

which one is that?

huh ok so i changed the samplers

and some of the make the issue worse, and some of them resolves it

[ruby yoke]

yup. that's expected

use the ones that don't cause the issue

[gloomy bolt]

are some of the samplers not compatible with ponyxl? or what's the issue?

[ruby yoke]

yes.

[remote cosmos]

I just use Euler a for everything Probably not the best method, but havent had any issues with it

[ruby yoke]

it's not. it's good with some models, will not work with others.

[remote cosmos]

Really? Huh I haven't encountered any models it inherently doesn't work with. I suppose its an evenutality

[proper swallow]

@ornate elk It seems like using script SD upscale seems to work fine but as soon as I use Ultimate SD upscale I get FileNotFoundError: [WinError 3]

[ornate elk]

Oh okay. Have your tried updating the extensions?

[proper swallow]

yeah didnt work. Im gonna try reinstall the extension

[stone orchid]

Anyone got a good gui that i can use with the latest python

[ornate elk]

Python 3.12 doesn't support any webui rn

[neon orchid]

I haven't tried it yet, I didn't want to ruin it yet.

[unreal anchor]

So, now I installed ComfyUI trough stability matrix, but for some reason no matter what prompt I pull Up, it stays at 0%, any idea why?

[spiral berry]

Has anyone I mean anyone been able to install and use stable diffusion on AMD cards with rocm and not direct ml

I tried all the ways on automatic 111 page for linux (installed linux mint and endeavor os) still can't get it to work

[ornate elk]

Hey, yes and I made some guides for that too

[ornate elk]

Oh your on linux

Can't help much with that sry

I use AMD on Windoes with Rocm and Zluda

[spiral berry]

So you got rocm running on windows?

Wasn't zluda shut down or something

Because amd sucks

[ornate elk]

Its still there

A fork of it

That also gets updates

[spiral berry]

Show me the guide bro

[ornate elk]

Here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[spiral berry]

Aight bruv thanks

[ornate elk]

No problem, if you have any questions about it, feel free to ask

[spiral berry]

Yea its 2:37 am rn so am gonna try tomorrow

[ornate elk]

Alright

[loud pond]

hi! so i want a addon or extension or something where i can input a bunch of character lora's or names and it will generate a random one each time

[craggy epoch]

I'll double-check that... didn't think adblocker would be an issue as I had it set to not block anything on my local network. In this case, I was using the IP address of the machine (I use --listen to access from my LAN on other devices).

[proper delta]

i'm trying to download stable diffusion right now, but it keeps saying that it can't find any checkpoints even though i put the .ckpt file where it's supposed to be? I thought it might have been the version difference at first, but v1-5 doesn't work either, not sure what to do

[spiral berry]

Hi, what happens if I already have zluda in my path and amd hip sdk installed on the first run?

The guide says to run the bat file and you will get an error, after that you install hip and add zluda to your path

[silent stream]

I am using Lobos theme, and when and disable card descriptions, the text gets an error

[ornate elk]

Thats no problem then

If you already have hip SDK it should work.
Depending on your GPU.
As some GPUs need custom rocm library files.

[ornate elk]

Hey can you show us the cmd log?

[spiral berry]

Its not using my gpu at all

[ornate elk]

What's in your webui-user.bat?

And did you git cloned the correct repo?

[spiral berry]

yes I gitcloned the stuff from the guide

[ornate elk]

Have you generated the first image ?

This can take 15-40 minutes

[spiral berry]

yes

first image

[ornate elk]

Check the vram usage of the GPU when generating

Because you won't see much in the processing tab

[spiral berry]

rn its 6.6/16gb on idle

[ornate elk]

And when generating?

Also how long does it take?

[spiral berry]

It took a few seconds

512 x 512 image

[ornate elk]

Ok then its using your GPU

But you should increase your Windows Pagefile.
As you only have 16gb of RAM

[spiral berry]

I am running into not enough memory when generating stuff now

it doesnt let me generate

[ornate elk]

What settings did you use? Resolution, model, steps etc

And what's your GPU?

[spiral berry]

Rx6800

[ornate elk]

What's the model you tried?
Also make sure you don't run any Games or Wallpaper engine in the background

[spiral berry]

Dreamshaper_8

why is it using my ram shouldnt it use my gpu vram?

[ornate elk]

Its using your vram

But it also uses RAM

Make sure your on the latest AMD Driver

Also do you have hip SDK 6.1 installed?

Also can you show me a CMD log after you generated an image?

[spiral berry]

yes

[spiral berry]

where do I get that?

[ornate elk]

Copy the text of the cmd

The whole text with ctrl+a

[silent stream]

When you run the webui-user.bat, it opens a command prompt

That command prompt contains all verbose logging informations

[spiral berry]

I see I just closed it atm

[spiral berry]

by how much and on which drive?

I have windows on a sata SSD and I have a 1 tb nvme I use for games

[ornate elk]

Set the pagefile on the nvme drive and disable it for C.
Then set it to Min 16000 and Max 24000

Also make sure you always launch the webui-user.bat and not the webui.bat

And share the cmd log when you generated again

[spiral berry]

Thanks

It did generate this time

[ornate elk]

You have the wrong python version installed.

You need to upgrade to python 3.10.11 64bit

Your on Python 3.10.0 which isnt supported by any webui

[spiral berry]

Oh

How is it working then?

[ornate elk]

Its buggy working

Some stuff won't work and it causes issues

[spiral berry]

how do I change versions correctly?

[ornate elk]

You can just install python 3.10.11 64bit and overwrite the other version. No need to uninstall it first.
After that you have to delete the venv folder and relaunch the webui-user.bat

[spiral berry]

Okay I will try that

[silent stream]

Still having some issues with this

Seems to be somewhat random when it appears for me whenever the program restarts

Seems like some kind of language issue?

[ornate elk]

You better make an issue on the lobe theme github repo

[silent stream]

Hmm

[ornate elk]

Also make sure the webui is whitelisted in any browser adblocker

[silent stream]

I think it is possibley caused by adblocker like you said

Insane suggestion. It is gone for now, and will see if it comes back

[ornate elk]

alright np, yea adblockers can block css stuff too and lobe theme is heavily customized

[silent stream]

Also recently learned that the lycoris extension has been sabotaging my a1111

Like holy shit

Made PonyXL generation about 30 times slower

[ornate elk]

Yea lycoris extension is not needed since 1.7.0 xD

[silent stream]

Used to take 2 minutes

Now it takes 5 seconds

What the actual f-

[ornate elk]

Yea it manipulates some stuff in the background

[silent stream]

I suspected SOMETHING was screwing with it

Since CUDA usage was capping at around 25%

[ornate elk]

Ah ok

Also make sure your other extensions are updated

And make sure you use xformers for the best performance

[silent stream]

I use --opt-sdp-no-mem-attention --api --cors-allow-origins null

I heard xformer doesn't work with opt-sdp-no-mem-attention

Or am I experiencing another troll

[ornate elk]

Wouldn't use those 2 arguments

Xformers could be faster and uses less vram

[silent stream]

It is FASTER than opt-sdp-no-mem-attention?

[ornate elk]

Would need to test again but last time xformers was as fast as opt sdp attention

But uses less vram

The non mem thing is only useful if you have enough RAM

[silent stream]

I have 32GB RAM with 4090

[ornate elk]

Ah okay then it doesn't should make any difference

With a 4090 and 32gb you can use either or the other ^^

[spiral berry]

How do I use this model? Do I just drop it where I drop my checkpoints? https://civitai.com/models/372465/pony-realism

[silent stream]

Yes

Use it like any other checkpoint models

[spiral berry]

When I did that its downloading a 10 gb file

open_clip_pytorch_model.bin

[warped canyon]

is there any way to get the forge inpaint section more like a1111? I hate that the image gets zoomed out when i scroll

[ornate elk]

That should be one time only

Have you updated your python?

[fierce hollow]

A question for you.
Stable Diffusion is able to draw realistic images, cartoon, generally different styles, er otic, and what is the problem with drawing images with cruelty? (severed limbs, for example?)

[somber turret]

(Newest version of ForgeUI) Whenever I use Hires. Fix, I always get disfigured characters with any model I use. any suggestions or known issues with this?

[versed imp]

Anyone can tell me about what is this exactly please?

[ornate elk]

Make sure to set the denois to 0.5 or below and hires steps to 10

[ornate elk]

Would need the full cmd log

[versed imp]

This appears when i open SD, so i need to open again and print everything?

[pallid quarry]

Kind people, please help, here I have an image of a character, I need to make it into a T-pose, I installed a mod for lora, for a T-pose, it fits into a line, but the character does not get into a T-pose, I went to img2img in inpaint, nothing works

[ornate elk]

Yes

[open fox]

Did you tried controlnet open pose model?

[versed imp]

I have installed here, bc i want to learn how to make animated videos

[spiral berry]

Yes

Everything works now

I generated several images without any issues

[pallid quarry]

No

[open fox]

Short video I found with google that might explain the process.

https://m.youtube.com/watch?v=fJsi-swa5ZY

not mine just describes the process with auto11111

[pallid quarry]

thank you very much, I will definitely look at it

[spiral berry]

is there a guide to this thing?

because am overwhelmed

idk whats possible

Like whats inpainting

img2img

etc

what are the different parameters

[pallid quarry]

the video shows the example of the generated images, but there is one description for my image, and if you try to generate it, we give out many different options, but not my content

[ornate elk]

Perfect

[pallid quarry]

Simply put, I have pictures, these are screenshots from videos or posters

[ornate elk]

Okay, can you share the cmd log? Then I can check for any errors

[versed imp]

I think is this

[ornate elk]

Thats not the full log

You can use ctrl+a to get everything

[open fox]

So you used the reference controlnet and the open pose controlnet after using a AI-Model to create a prompt for the screenshots you got and it does not work?

[pallid quarry]

I tried, and there half of the body seemed to stand up as I want, but half of the face is missing, the description of the photo is not the most accurate, just the 50/50 method

[spiral berry]

How do I use this? https://civitai.com/models/721039/retro-anime-flux-style?modelVersionId=806265 it doesnt appear in Loras tab

[loud pond]

Any one have any ideas?

[proper delta]

this?

[ornate elk]

Okay, whats in your stable-diffusion-webui/models/stable-diffusion folder?

[ornate elk]

Ah thats an old checkpoint. Delete it and use this for example:
https://civitai.com/models/4384/dreamshaper

[proper delta]

ahh okay! ill download it and see if that fixes the problem

hm, still getting this error

[ornate elk]

Dont rename the model please

Also can you show me the full cmd log?

[proper delta]

do you mean the file /Users/.... model.ckpt thing?

[ornate elk]

Are you on Linux?

And AMD ?

Can you show what's inside the models/stable-diffusion folder?

[proper delta]

not sure, im on MacOS

[ornate elk]

Did you used the Stable diffusion install script from the pinned messages?

For mac

This:
[MAC] Installing Automatic1111 on Apple MAC OS:
https://github.com/viking1304/a1111-setup

[proper delta]

ah i used this video https://www.youtube.com/watch?v=Jh-clc4jEvk

should i delete the folder and start again?

[ornate elk]

Yep, use the script from the link

It sets up everything correctly

But keep the Dreamshaper model

[proper delta]

i see, ill do that then, thank you

[ornate elk]

No problem

[pale hatch]

I'm having the following error:

Collecting scikit-image==0.21.0 (from -r requirements_versions.txt (line 31))
  Using cached scikit_image-0.21.0.tar.gz (22.7 MB)
  <...>
  Preparing metadata (pyproject.toml): started
  Preparing metadata (pyproject.toml): finished with status 'error'

stderr:   error: subprocess-exited-with-error

  Preparing metadata (pyproject.toml) did not run successfully.
  exit code: 1

  [12 lines of output]
  + meson setup C:\Users\<user>\AppData\Local\Temp\pip-install-_asivvwm\scikit-image_1d25bf7b459d4076a87e73ed62e2d8d7 C:\Users\Strawb\AppData\Local\Temp\pip-install-_asivvwm\scikit-image_1d25bf7b459d4076a87e73ed62e2d8d7\.mesonpy-rv35rolv -Dbuildtype=release -Db_ndebug=if-release -Db_vscrt=md --native-file=C:\Users\<user>\AppData\Local\Temp\pip-install-_asivvwm\scikit-image_1d25bf7b459d4076a87e73ed62e2d8d7\.mesonpy-rv35rolv\meson-python-native-file.ini
  The Meson build system
  Version: 1.5.2
  Source dir: C:\Users\<user>\AppData\Local\Temp\pip-install-_asivvwm\scikit-image_1d25bf7b459d4076a87e73ed62e2d8d7
  Build dir: C:\Users\<user>\AppData\Local\Temp\pip-install-_asivvwm\scikit-image_1d25bf7b459d4076a87e73ed62e2d8d7\.mesonpy-rv35rolv
  Build type: native build
  Project name: scikit-image
  Project version: 0.21.0

  ..\meson.build:1:0: ERROR: Compiler clang cannot compile programs.

  A full log can be found at C:\Users\<user>\AppData\Local\Temp\pip-install-_asivvwm\scikit-image_1d25bf7b459d4076a87e73ed62e2d8d7\.mesonpy-rv35rolv\meson-logs\meson-log.txt
  [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed
Encountered error while generating package metadata.
See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.
```How do I fix this? I tried ChatGPT but it didn't know either.

[ruby yoke]

Compiler clang cannot compile programs< tracing that back - Preparing metadata (pyproject.toml) did not run successfully - and tracing that back -Preparing metadata (pyproject.toml): finished with status 'error' - so what's in your .toml file?

[pale hatch]

The normal one from the webui:

[tool.ruff]

target-version = "py39"

[tool.ruff.lint]

extend-select = [
  "B",
  "C",
  "I",
  "W",
]

exclude = [
    "extensions",
    "extensions-disabled",
]

ignore = [
    "E501", # Line too long
    "E721", # Do not compare types, use `isinstance`
    "E731", # Do not assign a `lambda` expression, use a `def`
    
    "I001", # Import block is un-sorted or un-formatted
    "C901", # Function is too complex
    "C408", # Rewrite as a literal
    "W605", # invalid escape sequence, messes with some docstrings
]

[tool.ruff.lint.per-file-ignores]
"webui.py" = ["E402"]  # Module level import not at top of file

[tool.ruff.lint.flake8-bugbear]
# Allow default arguments like, e.g., `data: List[str] = fastapi.Query(None)`.
extend-immutable-calls = ["fastapi.Depends", "fastapi.security.HTTPBasic"]

[tool.pytest.ini_options]
base_url = "http://127.0.0.1:7860"

[ruby yoke]

that's where the issue is, however - is scikit downlevel?

what is the log skipping at the <...> line here: Using cached scikit_image-0.21.0.tar.gz (22.7 MB)
<...>
Preparing metadata (pyproject.toml): started

[pale hatch]

Collecting scikit-image==0.21.0 (from -r requirements_versions.txt (line 31))
  Using cached scikit_image-0.21.0.tar.gz (22.7 MB)
  Installing build dependencies: started
  Installing build dependencies: finished with status 'done'
  Getting requirements to build wheel: started
  Getting requirements to build wheel: finished with status 'done'
  Installing backend dependencies: started
  Installing backend dependencies: finished with status 'done'
  Preparing metadata (pyproject.toml): started
  Preparing metadata (pyproject.toml): finished with status 'error'

Had to cut it out because of Discord's character limit

[ruby yoke]

okay, so it's specifically having an issue with creating the metadata - you can see that it starts working on it and then errors

so that's going to be a post on one of the python support forums i think

[pale hatch]

Hm, okay. What should I tell them? Like "My toml isn't preparing metadata, here is the error log"?

[ruby yoke]

post that text, exactly like you posted it here and ask 'can someone help me track down why it's having an issue with the metadata

[pale hatch]

Thank you!

[ruby yoke]

welcome

[ornate elk]

Yes

[ornate elk]

What's your python version?

Do you have an example? With your txt2img settings?

[pale hatch]

Ok this is weird, I specified it to set up an archaic version (that 3.10.6 version) but I just checked and it made it a 3.12.3 version. How do I fix this? Last time I downgraded my Python, it broke every program I made, plus my VS/C.

OOF, I have to downgrade?! The pain...

[ornate elk]

An example image

And the settings like resolution, model etc

[ornate elk]

3.12 is not supported by any webui

Better install 3.10.11 64bit

You can define the Path in the webui-user.bat

[pale hatch]

Do you mean webui-start.bat?

I remade the venv into 3.10.6 since setting it to the python path seemingly didn't work. Am running it now, and will get back if I have any further issues.

[proper delta]

btw, do i have to run ./a1111-setup.sh everytime i want to open stable diffusion? just wondering because it takes a while to set up since it seems to need to redownload torch, numpy, pillow and etc everytime

[ornate elk]

Nope you need to launch the webui-user.sh file

[proper delta]

ohh okay

[ruby yoke]

inpaint it out

[proper delta]

after using a1111-setup.sh to set up stable diffusion, the checkpoint/model seems to be v1-5 even though in the actual folder i've put the dreamshaper model

[ornate elk]

have you deleted the old webui version?

also try pressing the blue refresh button

[proper delta]

yep, i redownloaded a new stable-diffusion folder from github

[proper delta]

still only one option, v1-5

[ornate elk]

that makes no sense 😮

its not the correct webui folder maybe

[proper delta]

i only have one stable-diffusion-webui folder though

does this mean anything?
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled
Loading weights [6ce0161689] from /Users/reinal/stable-diffusion-webui/models/Stable-diffusion/v1-5-pruned-emaonly.safetensors
Creating model from config: /Users/reinal/stable-diffusion-webui/configs/v1-inference.yaml
Running on local URL: http://127.0.0.1:7860

[ornate elk]

where is it located?

[ornate elk]

nope

check this path please:
/Users/reinal/stable-diffusion-webui/models/stable-diffusion

[ornate elk]

thats the wrong folder

its not located in downloads

you can delete it there

[proper delta]

oh huh

[ornate elk]

its here:
/Users/reinal/stable-diffusion-webui/

[proper delta]

how do i put the dreamshaper file in? using terminal?

[ornate elk]

find the folder and put it in

best trick is to generate an image and then click on the folder icon below

then your at the right path and then go back to the models folder

[proper delta]

wait how do i find the folder if it's not in downloads? im not sure how to look in reinal

[ornate elk]

like i said, generate an image

then click on the folder icon

[ornate elk]

and now do you have a file path at the side or top ?

can apple do something like this xD

[proper delta]

OH i see it

thanks!

[ornate elk]

perfect, np 🙂

you should delete the stuff from the downloads folder to not get confused

and then maybe make a desktop shortcut to the webui-user.sh from the correct webui

[proper delta]

yep, already got rid of that one

[sterile garnet]

Does anyone know the latest compatible Graphics Driver Version for NVIDIA that works with CUDA 11.8?

[loud pond]

If I wanted to make stable diffusion generate a random character each time would I have to use a wild card with the Loras??

[loud pond]

Or is it just not possible

[ornate elk]

the latest one works

[steady thistle]

I did a series of in-painting on this image and the quality went from the bright at first to the dim image. I'm using a ComfyUI Flux workflow, Is there a way to prevent this image degradation?

[sterile garnet]

yeah because it wont let me launch comfy UI because it says my graphics driver is too old

[ornate elk]

ah okay, then just update to the latest and your fine ^^

[sterile garnet]

alr ty

[ornate elk]

np

[steady thistle]

I'm using venv on my ComfyUI install and I'd like to try Forge also for different tasks. Do I need a venv for the Forge install as well?

[ornate elk]

forge will create its own venv

[steady thistle]

Oh thank you, that's a relief.

Should I try Forge? Your install says there will be some change that will prevent daily use. Should I go to A1111 instead?

[ornate elk]

forge has two versions, a new and an old one

you can try the new one if you want

the old one is only for people who want the old version

so they should not update that version

[steady thistle]

Okay, I'll try the new one since this is my first try with Forge. Appreciated

[ornate elk]

made it more clear now what is the new and the old version thx

[loud pond]

just incase if someone has the same problem later one ive found out how to do it. You'd have to create a wild card with all your characters on it and it may include the loras, then you can generate a random character each time

[hasty maple]

@ornate elk Thank you for your pinned guide for installing SD nvidia or amd. I'm not seeing section for cpu. Im using Swarm ui. how do I ensure my Amd cpu does not take over for my gpu and make it strictly gpu bound. I check my task manager and use nvitop and see usage is high on gpu and ranging about 19 vram. Setup 3090 ryzen + 5950x

[ornate elk]

Hey, np.
It won't switch over to your CPU or cpus igpu because the webuis for nvidia work only with nvidia. They wouldn't launch otherwise.

Some extensions/Custom nodes may use CPU for some stuff.
But you won't notice that much

[loud pond]

@ornate elk if do i have to use the actual lora itself or can i just use the trigger word for the same effect?

[ornate elk]

You need to select it

And the trigger word can strengthen the effect as well as the kora strength number

[hasty maple]

downloaded FP8 from xlabs (Not quite sure the best one to get since black forest does not have FP8). My FP8 will not generate images I keep getting black image, but my FP16 generates images. The swarm ui debug log shows unet missing, but the files are in the respective locations.

[loud pond]

alright thanks!

[fickle panther]

I keep getting this error when trying to train an SDXL lora, never had this problem with training 1.5

Using kohya on a 4090 so vram not a problem

[ruby yoke]

start with this https://discuss.pytorch.org/t/finding-source-of-nan-in-forward-pass/51153/3 - the error is NaN detected in latents

[umbral garnet]

can someone help me out

I added this

found it on git but still not working

[mint chasm]

message# i make video error pop up plz help

[vocal burrow]

what's your gpu ?

[mint chasm]

nivda 1080ti

[vocal burrow]

wasn't asking you ^^". In your case you're probably missusing some node. can't help without seeing the graph

[mint chasm]

i do in stable swarm

Invalid operation: ComfyUI execution error: 'NoneType' object has no attribute 'tokenize'

[balmy lily]

is there any reason to use the standard comfyui installation instead of the comfyui thats on stabel swarm

[mint chasm]

comfyui also error

any one help me

[vocal burrow]

you re not using an SVD model with SVD_img2vid_Conditioning node

[mint chasm]

i use this model

[vocal burrow]

no ?

[mint chasm]

but not use this model

how i change

[vocal burrow]

click on it

or use the arrows next to it

[primal pendant]

hey how can i get automatic1111 to use same models checkpoints the others use already it's not in the system options

this ain't working manually adding in script what am i missing

If you want to link all models from an other Webui you can do that by adding --ckpt-dir "Path/to/models/stable-diffusion" --lora-dir "Path/to/models/lora" --controlnet-dir "..." --embeddings-dir "..." to the webui-user.bat, at the line set COMMANDLINE_ARGS=

[primal pendant]

nevermind figured it out lol

[steady thistle]

@ornate elk I'm looking to install Facefusion with venv rather than conda. Will this be the proper steps to do so?
git clone https://github.com/facefusion/facefusion
python -m venv venv
call venv\scripts\activate.bat
python.exe -m pip install --upgrade pip
python install.py --onnxruntime cuda --skip-conda
pip install --no-cache-dir --ignore-installed --force-reinstall --no-warn-conflicts torch==2.4.1+cu121 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

[ornate elk]

hey,yes this will work

[steady thistle]

Yay, I'm slowly learning.

[versed scaffold]

Been grinding for 3 days now!! Please help! This should be easy for someone else probably. My Reactor node won’t pop up when I right click. I got it working before but when I did the Reactor Face swap wouldn’t pop up either😭😭😭

I also see these issues and I have YouTubed and downloaded many fixes but nothing works🙈🙈

[glossy pumice]

Can I edit the bat file after it's downloaded everything

@ornate elk if you're here

@ornate elk help!!

Urgent

@vocal burrow

Csone

[steady thistle]

You can, and that's what I do.

[glossy pumice]

No no

I put some stuff on my bat file I didn't think it would work

What the hell is open clip and xfromers

@ornate elk

@vocal burrow

[steady thistle]

I deleted my request, I think I figured out what I did wrong.

[glossy pumice]

Good for you

[steady thistle]

Or maybe not. I might need to do pip install tensorrt==10.4.0 --extra-index-url https://pypi.nvidia.com in addition to the rest.

[ornate elk]

Yes

[steady thistle]

I just realized it was pointing to cu121 instead of 124, could that be my numpy issue with facefusion? I'm also installing the tensorrt and hopefully that fixes it.

I don't understand why i'm not getting a run.py either

[glossy pumice]

Really?

Nothing will happen

Even if I remove xformers

Or add no half

@ornate elk

[steady thistle]

I'm installing anaconda and it wants me to set miniconda3 as default python 3.12, but will that break my other python uses with ComfyUI?

[umbral garnet]

rx 580

[ornate elk]

If you remove xformers it will be slower

But yea you can add and remove stuff as you like

It won't break stuff

[ornate elk]

124 is not supported by many ai stuff

But idk if FaceFusion likes 121 more or if they support 124

[vocal burrow]

cf https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides#amd-automatic1111-with-zluda

[steady thistle]

Turns out facefusion doesn't require torch at all. The issue was where venv was installed. I did the following:

cmd in correct folder
python -m venv facefusion-env
facefusion-env\scripts\activate.bat
git clone https://github.com/facefusion/facefusion
cd facefusion
(check for pip upgrade)
python install.py --onnxruntime cuda --skip-conda
pip install -r -requirements.txt
python facefusion.py run --open-browser

Turns out I missed the pip install-r -requirements.txt

[rugged latch]

hai techsupport, im havin problems with the vacuum florescent display on my intrument panel, on the captains side, next to the PFD, on the DC-9 MD82 varient, its supposed to come on with the autopilot, to show config info, and is not happening, i think the display may be broken, any wya to fix it

XD

[steady thistle]

Did you try unplugging it and plugging it back in?

[rugged latch]

ah, negative, its screwed into the panel

[thorn badge]

noob question! will adding more RAM to my PC help run Flux, or is it GPU based RAM I need?

[remote cosmos]

Its VRAM you're looking for, past having enough RAM to run things well in general, system RAM won't help a ton.

[thorn badge]

ty. I have 10GB on my 3080.

[remote cosmos]

Lol wish I had one of those

But yeah outside of getting a new card there isnt a ton you can do hardware wise

[hardy void]

Hi. Do you need Lora to be train only one thing at a time?

Like If I want my Lora to prompt my hairstyle name, and my character name, then also I want it to remember my body too, like I show it with nake picture for realistic. So I can have fun with clothes swap while retain my own body shape.

Will that be possible?

[rugged latch]

boop

so erm

stable uses the maroon, and not the pink

[ornate elk]

Use a smaller flux model and make sure to have 32gb ram

And if you have only 16gb ram you need to increase the Windows Pagefile

[vocal burrow]

that's normal. shared memory is ram.

[rugged latch]

me as asking

sry for lack of question mark

xd

buh oki

i get it

is there a way to make stable use ram

or nu

[vocal burrow]

--use-cpu all. But you don't want to do that

[rugged latch]

oop, got u

[vocal burrow]

or be ready to one per output

[loud pond]

how do i install https://civitai.com/models/147759/remacri?modelVersionId=164821

i cant find a file called ESRGAN?

[proper delta]

if i want to switch from a1111 stable diffusion to forge, do I just rename the forge folder to stable-diffusion-webui and delete the old one?

[ornate elk]

you dont need to rename the folder

[proper delta]

oh does it just work automatically

[steady thistle]

Hey this is cool, you mentioned about the port being the same for facefusion and ComfyUI and so couldn't be run at the same time. So with some help I changed the port and can run both simultanously. It was pretty easy.
call D:\Ai\FaceFusion\facefusion_env\Scripts\activate.bat
set GRADIO_SERVER_PORT=7861
python facefusion.py run --open-browser

[balmy monolith]

I'm trying to run stable diffusion webui, but get an error; 'can't run without a checkpoint. Find and place a .ckpt file...'

nvm, i need a model.

[loud pond]

@ornate elk do you know where i can find this file?

[ornate elk]

its in the upscaler tab when you enable hires fix

RESRGAN 4x

[rugged latch]

look at all these deleted msgs

ooo

like ur Ui there bob

mine is like..

Yes

[loud pond]

oh alright thanks

[silent stream]

Best way to have Stable Diffusion WebUI and Forge share same model,loras, etc?

[rugged latch]

i dare

sum1 to make a batch file
type in the batch file
:loop
start
goto loop

then, save as batch.bat, then double click it

[primal pendant]

how do i fix connection errored out reguardless of what i try to create

was trying forgeui with the flux nf4 v2

i deleted venv still does it after rebuildign

[solemn atlas]

Hello, I now unplugged my ssd. Should I delete everything and install windows again or what should I do?

whatever I try it doesn't work

[snow breach]

does Forge UI support Flux? As soon as I click Generate, she explodes with no error message and kills the server.

[primal pendant]

i honestly don't think it does not for amd atleast with zluda

i can't get any working

all would be so much easier if i had nvidia lol but i hate team greed for value

[snow breach]

Well I've got a 4080 Super and I get the above prob. I'm guessing its me and not the card though!

[primal pendant]

did you delete venv and let it rebuild ?

that fixes most issues apparently

[snow breach]

So I've noticed. No its a fresh install from a week ago, everything else(SD15, SDXL) is fine.

[ornate elk]

Nf4 won't work with AMD

You have to use either the fp8 16gb model or the gguf models

[ornate elk]

You may have the wrong flux model

[ornate elk]

Try the 16gb fp8 from here or the nf4 flux.
https://huggingface.co/Comfy-Org/flux1-dev/tree/main

[solemn atlas]

I tried everything but I can't do anything after removing my ssd

[ornate elk]

so only the drive with windows is now connected?

the small drive or the 4tb drive?

[solemn atlas]

I unplugged my ssd with 4Tb and now only my computers ssd is in there because I accidentally clouded my c drive to d drive.

My PC has no windows but now I can't even delete everything or start everything from new

[ornate elk]

you have to use an USB with windows on it then you can install it on the PC from that

[solemn atlas]

I have. I put that USB in there but it doesn't show anything.

[ornate elk]

you hav to go into the boot menu

not into uefi

when restarting the PC, press F9 or F10

the whole time

[solemn atlas]

I'll try

[ornate elk]

it could be also F12

@solemn atlas also checkout this video to see the whole steps:
https://www.youtube.com/watch?v=G5TGARdHkRM

[solemn atlas]

thank you I will do it.
But I also have windows on my ssd 4tb. does that matter if I plug it again?

[ornate elk]

yea you have to plug in the drive while the PC is running. Then check if the drive is appears in the File Explorer
Then it wont use the windows from the 4tb. and then you should be able to delete the Windows Folder thats on there

[snow breach]

Damn, maybe should have bought the 4090 after all instead of the 4080 Super. That doesn't work either. Got a little further though. Attempted to free 16064MB for Text2Img and 18500MB for Img2Img.

Screen goes black and it crashes out haha.

Should I try the GGUFs?

[ornate elk]

the 16gb model should work. did you used comfyui or forge?

[snow breach]

Forge

[ornate elk]

make sure to not set the vram weight to max in forge

also dont use the t5 text encoder when using the fp8 model

only the model and the vae

[snow breach]

alright let me give that a shot. it complained about t5 missing last time so I didn't change it for this one.

ok that sort of works. thanks for that. text to image is great, but img2img is total trash. back to sdxl!

[solemn atlas]

is this right?

[ornate elk]

Thats the BIOS. You can move the two USB symbols at the top to the right so they are primary boot devices and them restart

[solemn atlas]

you mean the USB Symbols? but there are 4 of them

[ornate elk]

Only one of them is a USB stick

Move that to the right

[solemn atlas]

oh okay like this? and then?

[rugged latch]

press the easy button

[ornate elk]

To the left sry

Not to the right xD
Then save and reboot (F10)

[rugged latch]

remember when BIOSes looked like dis

those were the days

[solemn atlas]

ah ok haha so like this?

[rugged latch]

lezz not forget this

[rugged latch]

dang, yall tryna set boot order

i remember when it was as simple as, using up and down keys

to set it

[solemn atlas]

with the new design I don't understand anything haha

[ornate elk]

Now press F10 and save and reboot

[rugged latch]

yas xd

actually wait nu, it was plus and minus

[solemn atlas]

now all I see is this.

automatic repair

[ornate elk]

Okay then reboot and press F9, F10 or F12

To get into the boot menu

Not bios

Then select your USB drive

[solemn atlas]

is this the boot menu?

[rugged latch]

ya

u got 2 options for boot device select looks like

naviagte to the one that is ur USB

and press return

[ornate elk]

Yes use the intenso

[rugged latch]

imagine it says

BOOT FAILED
please insert a bootable disk and press enter

[solemn atlas]

it doesn't work

[rugged latch]

oofta

[ornate elk]

Then you didn't prepared the boot stick correctly

It has to be made like in the video i linked you

[rugged latch]

sry i jinxed u

[solemn atlas]

then I need to plug in my ssd again to do it again right?

[ornate elk]

Yes

Or on any other PC

[solemn atlas]

does it work on a surface too? so send it to my usb

I plugged my ssd but it doesn't worj

work*

just shows the same blue window

[steady thistle]

I'm having issues with my Facefusion install using venv. All the docs are for conda. If I install Anaconda and let it set itself as the default Python, is that going to break my ComfuUI venv setup?

[rugged latch]

is it possible u dont have ur BIOS setup to boot UEFi

prolly shoulda checked befire tryna boot nu?

sum BIOSes have an option to boot legacy or UEFI partitioned OSs

so, ur stick may stll be oki '

jus may need to setup BIOS

if ur tryna boot UEFI on a legacy boot BIOS, it wont boot it

need change setting

tho, ig it is also possble that UEFI is default

[ornate elk]

If the surface has an USB slot you can use it too

[solemn atlas]

I use this to connect it

[rugged latch]

-dabs in 80p-

[steady thistle]

@ornate elk Can you advise me on installing Anaconda? It wants to be the default Python, will that hurt my venv installs somehow? Or going to cause me some unexpected issues later on?

[ornate elk]

Why you want anaconda?

[primal pendant]

@ornate elkok cool so i can use one then which is best ?

[steady thistle]

Because Facefusion is an install with conda and I'm having issues with having it recognize CUDA and use it. The support for Facefusion seems to think it's because of using a venv

[ornate elk]

Gguf uses less ram and vram so would use that

[primal pendant]

do you have a link ?

[steady thistle]

This is the issue I'm getting.

D:\Ai\FaceFusion\facefusion>call D:\Ai\FaceFusion\facefusion_env\Scripts\activate.bat
Running on local URL: http://127.0.0.1:7861

To create a public link, set share=True in launch().
[FACEFUSION.CORE] Processing step 1 of 1
2024-10-05 09:35:16.5538034 [E:onnxruntime:Default, provider_bridge_ort.cc:1992 onnxruntime::TryGetProviderInfo_CUDA] D:\a_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1637 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "D:\Ai\FaceFusion\facefusion_env\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

2024-10-05 09:35:17.3148282 [E:onnxruntime:Default, provider_bridge_ort.cc:1992 onnxruntime::TryGetProviderInfo_CUDA] D:\a_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1637 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "D:\Ai\FaceFusion\facefusion_env\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

[FACEFUSION.CORE] Copying image with a resolution of 1536x1536
[FACEFUSION.FACE_SWAPPER] Processing
[FACEFUSION.CORE] Finalizing image with a resolution of 1536x1536
[FACEFUSION.CORE] Processing to image succeed in 2.28 seconds

[ornate elk]

Here:
https://huggingface.co/city96/FLUX.1-dev-gguf/tree/main

[primal pendant]

nice thanks alot which one

so many

the 23.8 ?

[ornate elk]

Try this:
Go into the venv with cmd and run
The venv activate and then
Pip uninstall onnxruntime-gpu
Pip install onnxruntime

[ornate elk]

The Q5_0

If you have only 8gb vram then the Q4_0

[primal pendant]

nice even better less download

i got 12gb

[ornate elk]

Also here read the gguf part:
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1050

For Text Encoder go with t5-fp8

[steady thistle]

It was a very quick install but when I ran Facefusion there was no longer an option to run with cuda.

[ornate elk]

What torch version is inside the venv ?

[steady thistle]

Ah, it wasn't installed with torch.

[primal pendant]

ok thanks got them all except the vae i can't get it for some reason

[ornate elk]

Then run pip install torch with the recommended version

[ornate elk]

Here direct link to VAE:
https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/ae.safetensors

[ornate elk]

Ah well you only get it when logged in into hugginface

[primal pendant]

stupid vae

lol i need it

[ornate elk]

I can upload it if you want

[primal pendant]

please do

[ornate elk]

Oh wait here it is:
https://civitai.com/models/619150/flux1-dev-vae

[primal pendant]

its dead everwhere i try

this don't have the ae one

[steady thistle]

Do I need torchvision and torchaudio as well?

[ornate elk]

Yes

[steady thistle]

I did them all using cu121 but no luck. It's an older version I got from a reccomendation online but that might have been for Facefusion 2 rather than 3. What would be the command for installing the most recent version?

By no luck, I mean that cuda is not offered as an option on the interface

[solemn atlas]

now it worked but I get this

oh forgot to unplug my ssd

[ornate elk]

shut down the pc and unplug it

you dont want to lose the data on it

and make sure you have everything you needed moved to the 4tb as you gonna wipe your C drive

[solemn atlas]

I unplugged it while it was on.

[solemn atlas]

yes I send everything to my other USB Stick because somehow I couldn't move them to my 4tb

[ornate elk]

okay

you still need to reboot to only see the small ssds partitions

then you can select them and delete them.
At the end you should have only 1 partition left. then click on next

[primal pendant]

is your windows that broken where your only choice is clean install ?

[ornate elk]

let me try getting it to work on my old pc

[steady thistle]

I appreciate it, I have to run before long but I'll implement whatever you suggest when I get back

[solemn atlas]

long story.. I cloned everything wrong

[primal pendant]

oh right i see

btw does flux get quickler after first run ?

this gonna take ages

[solemn atlas]

idk what happend now it just doesn't show the installation. Only this

[primal pendant]

oh damnn

windows recovery always been trash

[ornate elk]

then your havent booted into the usb stick

you need to get into the boot selection again

[ornate elk]

you need to shutdown

[solemn atlas]

how does.. that work

[primal pendant]

can't you promt and try to fix ?

[ornate elk]

and again restart while constantly pressing f9 or f12 or f10

[primal pendant]

is using flux like running for first time or will it always be this slow

😴

[solemn atlas]

what should I do after it

[ornate elk]

you need to be on that menu and then jus tselect the usb stick again (intenso)

[solemn atlas]

So I just need to choose the first one

[solemn atlas]

but that's what I get

[ornate elk]

that means it skipped booting from the usb drive

[solemn atlas]

I don't understand it.. just before it worked and now it doesn't. I really don't know what to do

[ornate elk]

yea its strange

try again and make sure the usb drive is plugged into the back of the PC

not front pannel

[primal pendant]

zzzzzz

if it's gonna take 30 mins for one image forget that lol

[solemn atlas]

I try.. but what if that also doesn't work

[ornate elk]

then you need to redo the USB installation step with the windows tool

[solemn atlas]

ok it wont work. same problem like before

[solemn atlas]

how does that work?

[spiral berry]

Has anyone gotten flux running on their AMD card?

[ornate elk]

like in the video i linked, where he prepared the usb drive

like you did before

with the surface

[ornate elk]

yes

[primal pendant]

it kinda running on my 6700 xt takes almsot 30 mins for one image

[ornate elk]

how much RAM do you have?

[primal pendant]

32 gb

[ornate elk]

ok

have you tried the fp8 16gb model before?

not gguf

[primal pendant]

would it be faster ?

[ornate elk]

could be

[primal pendant]

cause 30 mins for one is ridiclious

[ornate elk]

yea not normal

[primal pendant]

i will try f8 then

where is that one same place ?

[ornate elk]

here:
https://huggingface.co/Comfy-Org/flux1-dev/tree/main

[primal pendant]

ok will fuck that other off then

don't have room for both lol

i will most likely get another 2 tb drive at some point or atleast a 1 tb ssd

jsut for ai stuff there so dam big

it will work in forge right ?

as i don't have comfyi yet

[ornate elk]

@steady thistle so i made a launch script for FaceFusion.
It will create a venv and install torch 2.3.1 cu121 in it.
With that it shows up on my old PC with GTX1080

Place the Bat inside the Facefusion main folder.
Make sure to delete the venv before running the script

[ornate elk]

yes it will work

[ornate elk]

maybe you should download WizTree and check whats taking up so much space xD
and maybe delete some Games you dont play etc

[primal pendant]

well 2 tb is music from 93-2024 for traktor

i will remove gta v

[solemn atlas]

okay, I installed it again on my USB Stick and this is what I got

[ornate elk]

okay. now you can click on every entry but NOT the 32gb one!
then hit Delete partition.
then you should only have 2 partitions left, one is 1tb (900gb) and a 32gb one (the usb stick itself.)
Select the 900gb file and click next

[solemn atlas]

So I should delete everything expect these two in yellow?

ok it didn't worked. it says" installation failed"

[ornate elk]

yes

[ornate elk]

okay delete the 900gb one too

then select the 900gb drive and click next

[solemn atlas]

so I need to delete windows from my USB Stick again and install it again right? because it doesn't work a 2nd time

[ornate elk]

try reboot again and maybe it will get recognized

[solemn atlas]

I tried many times but it won't work

[primal pendant]

seems just as slow

how long does it take you ?

[open fox]

i only could say you how long it takes on a 4900 but this won't help you i guess.
whicht t5 Model do you use?

[ornate elk]

hmm then yep redo the stick but its strange

[primal pendant]

haven't done anythign except the check point and something simple that shouldn't take long but it really is

💤

[ornate elk]

try again with a resolution of 768x768

[open fox]

Guess the vram is full and it starts swapping.... This hurts a lot the performance

Pretty sure CS1o knows a neat trick to start the ui with some additional parameters to reduce the usage furthermore

[primal pendant]

yeah only have 12gb vram

[open fox]

Did the smaller resolution speed up the process significant?

[open fox]

Still to high i would say.

[primal pendant]

nope lol

well it's not 30 mins

[open fox]

Well it still says you are missing some memory to run it. May i suggest to try a pruned version of Flux.Schnell

[primal pendant]

but still slow

oh ok sure can try another

[open fox]

https://huggingface.co/city96/FLUX.1-schnell-gguf/tree/main

I hope that webui is gguf compatible. I use comfyui so i am not sure

[primal pendant]

already done this one i think q5

that was slow to

[open fox]

biggest advantage of Flux.Schnell is that you would need only 4-5 iterations. The result is not as good as the Flux.Dev but you should be able to generate a lot more images in the same time and one these might be signifikant better.

[primal pendant]

oh right i see sounds nice so which one would you suggest for me with a 6700 xt ?

[open fox]

Would try flux1-schnell-Q5_K_S.gguf

[solemn atlas]

now I start the PC this opens

[solemn atlas]

now it's even weird than before

[primal pendant]

@open foxok will try that now then

[ornate elk]

you have to click f9, f10 or f12, whichever opens the boot menu, before the bios screen appears

you have to spam the keyes while the pc boots

[solemn atlas]

I do that everytime so i deleted win11 again and put it again in my PC and I get this. I click on this and nothing happens

[ornate elk]

nothing happens?

[solemn atlas]

no, I reboot 3 times

[ornate elk]

something is maybe wrong with the usb stick then

[ornate elk]

the thing is even without any drive you can boot into an windows bootable stick

[solemn atlas]

the USB Stick is fine. It's completely new. I ordered it just a few days ago

[ornate elk]

really strange

and if you click on enter setup your in the bios?

[solemn atlas]

yes then I'm in the bios.

I delate win11 again and install it again. I don't know for how long I need to do this.. I try to install that now for over 3 hours 😔

[ornate elk]

have you tried other usb ports?

at the back

[solemn atlas]

yes i tried

[ornate elk]

hmm

[ornate elk]

also read this:
https://de.msi.com/support/technical_details/MB_Boot_OS_Entry

and dont use legacy_uefi

[primal pendant]

can't you go back to windows 10 ?

i ain't going 11 untill i'm forced to

[solemn atlas]

read it but it doesn't help me at all. It won't work at all

[ornate elk]

cant help much other than that as you only need a bootable stick

with windows on it

[solemn atlas]

is that right?

[solemn atlas]

I have both. I mean it worked once a few hours ago but now it doesn't. I even bought a new USB stick only for that

now I dealted all those other partition things and unplugged my ssd.. now I can't even use it like it was before. Now nothing works.

[ornate elk]

put that on 1

[solemn atlas]

it's on one

[ornate elk]

and when you reboot nothing happens?

alos try plugging in the stick into the front pannel and reboot

[solemn atlas]

is that correct??

[ornate elk]

yes

[solemn atlas]

cuz the link u send says boot manager

[ornate elk]

should work, but can you remake the stick again to be sure?

[ornate elk]

i didnt looked correctly so you may have deletet the usb boot manager here by accidant

so better redo the usb stick and then try again

[solemn atlas]

I deleted everything you said. I even send a picture.. but yeah I try again

[ornate elk]

yea i know, but you deletet also a 500mb thing from the usb.
so next time DONT delete those two but all other ones.

these two are from the same usb stick

so only delete all that is labeled "Datenträger 0"

[solemn atlas]

and how to undo that?

[ornate elk]

you just need to make the usb drive again with the windows tool

[solemn atlas]

but that's what I did the whole time and it didn't worked. But I try again

[ornate elk]

it should work

[solemn atlas]

I hope

[solemn atlas]

I download it again and try it again

it doesn't work

[ornate elk]

it doesnt?

okay then you can try replug the sata 4tb sdd and boot from that see if that works

[solemn atlas]

it doesn't.. same as before.

[solemn atlas]

and what should I there? I plugged it in

[ornate elk]

does the pc boots with it ?

because then you could reformat the other SSD from there

[solemn atlas]

yes it worked. Some hours ago it didn't worked.

[solemn atlas]

but not my 4tb right?

[ornate elk]

okay do you see the other drive ?

[ornate elk]

yea dont format the 4tb

[errant thistle]

Okay so, on automatic1111, how does one use multiple character loras and not have those two characters merge together? Or what extension do i need?

[ornate elk]

you need to use the regional prompter extension to do that

[ornate elk]

can you open disk manager?

[solemn atlas]

do you mean this?

[ornate elk]

nope

i mean the disk management

[solemn atlas]

ahh so this

[ornate elk]

can oyu scroll down and show the stick too

[ornate elk]

seems like the stick isnt formatted correctly

i will reformat a stick myself to verify

try download an ISO file for windows 11.
you can do that by running the media creation tool and select ISO when it asks for ISO or USB

with that ISO and a tool named Rufus you can then recreate a bootable usb stick faster

[primal pendant]

hmmmm thought i would install comfyui having a issue the manger not showing for me

it's there

[solemn atlas]

ok i try

[ornate elk]

have you relaunched comfyui? what does the cmd shows?

[primal pendant]

hahhahah nevermind it comeup now it was being weird i closed and relaunched a few times

now just need to it use folder

am i blind i don't see a way lol

👨‍🦯

Linking Models, Loras, etc. from other Webui's or Folders to ComfyUI:
If you want to link all models from an other Webui you can do that by renaming the extra_model_paths.example.yaml to extra_model_paths.yaml
Then right click the file and edit it with Notepad or Notepad++.
If you have Automatic1111 installed you only need to change the base_path line like in my Example that links to the Zluda Auto1111 Webui:
base_path: C:\SD-Zluda\stable-diffusion-webui-amdpgu
Then save and relaunch the Start-Comfyui.bat

in the comfy bat ? i don't see it

[ornate elk]

okay tested it on mine and the stick looks correct

[primal pendant]

where is yaml ?

[ornate elk]

@solemn atlas you can try to delete this partition: (the red sqaure with the 100mb)
and make sure you delete it ONLY from the 900gb drive and NOT from the 4tb

[solemn atlas]

okay, i'll just do it

[ornate elk]

hmmm

okay then try the iso rufus boot stick method

[solemn atlas]

ok it's finished

[ornate elk]

Here is also a quick explanation on rufus and how to make the usb drive ready.
https://www.youtube.com/watch?v=NSRCZEKDMK8

[ornate elk]

yes i got the same error but the tool still works

it could be caused by not having cuda toolkit installed

or the wrong one for pytorch

[steady thistle]

I see, but the occlusion option is missing from the face debugger.

I needed that for vampire teeth and sticking out tongues. I know it's silly, but that's the only reason I was using the standalone vs faceswapper.

[ornate elk]

you could also try Roop-unleashed

it could work better than face fusion

[solemn atlas]

now that I thought it works i get this.

[steady thistle]

Okay, I'll look into that. I knew roop was depreciated so I didn't know to look for another fork

[ornate elk]

close the file explorer and any open application and try again

[solemn atlas]

"the drive is being used by another process. Please terminate this other process"

[ornate elk]

yea, the fork is quite active and works
https://github.com/C0untFloyd/roop-unleashed

[steady thistle]

Nice. That's great.

[ornate elk]

you can also restart the PC and then use Rufus again

that should work

[steady thistle]

Does this work in Forge? I installed in forge but don't know where to access it yet.

[ornate elk]

its not an extension

its standalone

[solemn atlas]

ok this worked

what should i do now

[ornate elk]

now plug in the stick at the back of the PC and try spamming F9 while booting

or shutdown and unplug the 4tb before

[primal pendant]

i can't get models to work at all do i have to have only in automatic11111 for it to work ?

[solemn atlas]

this is what I get when spamming f9

[ornate elk]

okay then spam the other F key that you used to get into the boot selection

when you reboot

maybe F12 or F10

[ornate elk]

yep select the usb drive

[solemn atlas]

i selected uefi but it doesn't work

[ornate elk]

thats so strange

[ornate elk]

hmmm

can you go into the setup and there make a picture ?

[ornate elk]

okay looks normal. can you enable ez mode at the top?

[ornate elk]

looks correct

i really dont know how you could fix that.
May last idea is to boot into the 4tb ssd windows. then there Format the 1tb ssd so that it loses its uefi and then try again to boot into the usb stick while the 4tb is not plugged in

[ornate elk]

oh one thing is you can check if fast boot and secure boot is disabled. that could help
set them both to disabled
its somehwere under the Advanced mode and then on settings#

and then on boot options

[solemn atlas]

you mean this?

after nearly 8hours of trying I'm also tired. Thank you so much for trying to help me. I would like to do the other thing tomorrow

[primal pendant]

can someone please explain to why the hell comfy ui don't show my models even after adding paths

C:\SD-Zluda\stable-diffusion-webui-amdgpu-forge\models

this is where they are

[ornate elk]

yes, no problem. yea try the other thing tomorrow and let me know

[ornate elk]

if you have automatic1111 installed you should move your models there. its easier to link them better to comfyui

[primal pendant]

yeah guess no choice as it don't matter what i try no models show

stupid thing

[ornate elk]

in my guides its explained how to link models from auto1111 into each other webui

[primal pendant]

all other ui's hasd no issue

but this thing just refuses

to show

[ornate elk]

thats how it looks for me in comfyui from auto1111

[primal pendant]

thats autonmatic though right

[ornate elk]

yea but for forge its the same

as forge is based on auto1111

[primal pendant]

my forge ain't working for comfyui that's where my issue is

[ornate elk]

remove the \models from the base path

[primal pendant]

oh that did it

lol

wow it don't like alot of checkpoints that forge had no issue with

the flux won't even show or xxl

seems only sd1.5 are showing this thing so trash

[ornate elk]

you need specific nodes for the models

a flux workflow wont work with sdxl or 1.5

[primal pendant]

i can't egt anythign to work lol

[ornate elk]

here is a workflow for flux fp8

here is a sdxl workflow

[primal pendant]

that don't matter can't even show a image right now lol

Numpy is not available

for anything i tried

[ornate elk]

ah this error

you may need to manual install numpy into the venv

[primal pendant]

hmmmm

[ornate elk]

someone fixed it with this:
Installed comfyUI manager, went to ComfyUI\custom_nodes\ComfyUI-Manager\config.ini and wrote security_level = normal-, restarted server, went to "manager", "install pip packages" and wrote numpy==1.26.4 even tho it was already installed

[primal pendant]

i do that inside the program or clsing and text doc edit ?

[ornate elk]

the first is doc edit

[primal pendant]

is says normal without me chnaging it

[ornate elk]

ok good

[primal pendant]

manager so inside the app ?

[ornate elk]

yep

click on manager and then on this

[primal pendant]

hahahah

so security off i guess?

and not normal

or set to low ?

[ornate elk]

try that

[primal pendant]

so false ?

or low

or none

even set to none it won't

[ornate elk]

okay then go into the comfyui manager folder

there click into the explorer bar, not searchbar.
then type cmd and hit enter

then run:
pip install numpy==1.26.4

[primal pendant]

that works atleast

lol

oh wonder of thats; why

added

nope

[ornate elk]

you need these two:

[primal pendant]

added long time ago

well about a week ago

i pretty much added all to path to do with python afetr i had those errors before

[ornate elk]

thats an other path as you added at the bottom

[primal pendant]

i don't see python in appdata

i assume local ?

[ornate elk]

hmm

[ornate elk]

yea

local/porgrams/python

[primal pendant]

i don;t ahve one

[ornate elk]

okay then
try go into the comfyui folder
there click into the explorer bar, not searchbar.
then type cmd and hit enter
then run:
pip install numpy==1.26.4

[ornate elk]

programs

its in the image at the bottom

[primal pendant]

yeah mine in programs

shall i move it ?

[ornate elk]

no

it has to be in there

[primal pendant]

ok i will give full control to

[ornate elk]

no

dont do anything with permissions

you gonna break stuff again xD

[primal pendant]

hahahaha

[ornate elk]

in the environment path

if that wont work then try what i said here:
#🤝｜tech-support message

if that wont work, idk. as im going off for today

[primal pendant]

comfyui is biggest piece of trash why do all others woirk except this

same error

[ornate elk]

With your GPU its also not worth using comfyui

As forge and auto are more advanced when it comes to reducing vram and RAM

[primal pendant]

still nothign lmao

[azure magnet]

How to fix it

[ruby yoke]

make sure you're using a lora created for the model you're using

[junior python]

hello i want to generate video to pic on next view but this happend How to fix it

[sly pike]

I've installed stable diffusion and tried to generate a picture but there is a problem: | RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx | I have cpu intel, can i use Automatic1111? is there something like automatic supporting CPU?

[rugged latch]

i mean

i dont have a card

i have AMD 5600G with vega 7

works for meeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee

maybe u need an AMD

||intel sucks anyway ||

||sodoes nivida||

[clever stream]

How do you use this software? Is there a tutorial for it?

[rugged latch]

redline, is where u select ur checkpoint and/or model, orange is where u tell it wut u want, yellow is where u tell it wut u dont want, blue circled is how much u want it to follow the prompt, word for word, white circle is where u press to make it do stuff

to use a "lora"
type this under ur prompt, in same box (orange) without the quotation marks ofc, "lora:lora_name:1wutever lora does"

the (1) after the semicoloan, represents how much u wan stable to use the lora u chose

0.1-1

[clever stream]

Thank you very much!

[rugged latch]

ofc =3

[ornate elk]

Go into appdata local temp and there create a folder named gradio

[ornate elk]

You can use the slow CPU mode if you want.

For that you need to edit the webui-user.bat and add:
--skip-torch-cuda-test --use-cpu all --no-half --precision full --opt-sub-quad-attention
At the line commandline_args=

[ornate elk]

For an install guide checkout the pinned messages of this channel

[sonic light]

does anyone know whats going wrong here?

it's in ComfyUI

[solemn atlas]

how do I boot into the 4tb ssd?

[sly pike]

I'm entirely grateful for you

[junior python]

still error, idk why i do that many time

[rugged latch]

is ur BIOS in uefi mode

or legacy

and also make sure secure boot is off

legacy mode wont boot uefi

uefi mode will boot uefi

[solemn atlas]

I think it's uefi

[solemn atlas]

how do I do that?

[rugged latch]

it be in bootmainetence manager in advanced boot options

i hate that BIOSes are all so feckin diff now xD

with all these GUIs, and settings buried in diffirent places

imp

shoulda jus stuck to making things standard

[loud pond]

Does anyone know why ReActor isnt generating images which have faces simular to the image inputted

[warped canyon]

SD Forge is not playing notification sound when process is complete (only plays if I'm already on that WEB-UI tab)

[loud pond]

also when using the upscaler why is it giving this error?

[fair oxide]

anyone knows how to make text look good ?

[loud pond]

on the generation?

[fair oxide]

in paint can be too