🤝｜tech-support

[tawny saddle]

boop'

[thorn barn]

Does anyone know how to best preserve lingerie details and fit when using AI to replace the model? I need to shoot the lingerie on a dummy model first, then replace it with an AI-generated human model.

[vapid bay]

Hey guys! If anyone could help me I would be so grateful. I've been trying to get Forge to work and it's starting to feel desperate.
The sole problem is that the Forge gallery doesn't show generated images. It only shows the recent one I just generated, and it always changes to the newest one. Only if I generate a batch, I can see multiple images in the gallery. And then that also disappears if I generate a new image.
I've been trying some Gradio fix in webui-user.sh, updated the UI, fiddled around with settings, disabled all extensions, reinstalled, but nothing.. I'm on a vast.ai instance if that matters and Forge worked just fine like 2 weeks ago when I gave it a go.

[mossy totem]

aight I'll try that now

[mossy totem]

my picture is just a grey box

I don't know what I did wrong

[tawny saddle]

uhh

i think stable is just saying it likes the colour grey, maybe its trying to hint that u should use it in sum art

xd

[mossy totem]

what does seed do?

[ornate elk]

You probably used a lora instead of a checkpoint to generate the image

Make sure the checkpoint (model) file is at least 2gb in size

[agile basalt]

Is it okay to ask questions here relating to comfyui?

[ornate elk]

Yep, but you can also ask in #🧣｜comfy-ui

[agile basalt]

oh didn't see it, thank you

[shadow slate]

i need help guys

why animate ui tab not showing i installed it correctly

[sudden pike]

@ornate elk i have a compilation is in progress. please wait loop when i wanna use SD again

i changed nothing

[ornate elk]

Can you show me a full CMD log?

[ornate elk]

That can happen when for example the AMD Driver uodates

[ornate elk]

Wait remove --skip-torch-cuda-test --no-half --precision full from the webui-user.bat
And only add --use-zluda --skip-ort --update-check

Then relaunch

[sudden pike]

also downgraded

should i do DDU?

[ornate elk]

No you can upgrade to the latest driver version if you want.
The dev updated the webui a few days ago so to support it

But you need to update the webui too.

With git pull

[sudden pike]

doesnt want to start with 25.6.1

[ornate elk]

Because its not updated..

[sudden pike]

idk how to

[ornate elk]

Close the cmd

Then edit the webui-user.bat
In the second row/line
Just add
git pull
Then save and relaunch

[sudden pike]

Already up to date

[ornate elk]

Okay then it should work

[sudden pike]

nope

[ornate elk]

Which driver version are you on RN?

[sudden pike]

25.4.1

[ornate elk]

Okay then it will work

[sudden pike]

but it doesn't

[ornate elk]

It compiles

Give it time

It will work

After thats done

[sudden pike]

its spaming my console with that

[ornate elk]

Yea thats normal

[sudden pike]

is it a one time thing?

[ornate elk]

It will compile only if torch, zluda or amd driver updates

Or changes

So when its done it won't compile again

[sudden pike]

how long does it take?

[ornate elk]

15-40 mins max

[sudden pike]

bruh okay xD

should i go to 25.6.1?

[ornate elk]

I would wait

I'm still on 25.3.1 with my 7900xtx

[sudden pike]

ah ok ok

i wish AMD has better vram management

[ornate elk]

What's your GPU?

[ornate elk]

With zluda its quite good

With directml it was horror

[sudden pike]

ah it works now

ok thanks
me impatient mf lol

[ornate elk]

No problem ^^

[shadow slate]

@ornate elk

[ornate elk]

What's your GPU?

[shadow slate]

rtx 3050

[ornate elk]

Okay. Edit the webui-user.bat
Remove --skip-torch-cuda-test
Then add --xformers
Save and exit.
Then delete the venv folder.
Open up a cmd and run
Pip cache purge
Then relaunch the webui-user.bat

[shadow slate]

like cmd should be in admin mode?

[ornate elk]

Nope

[shadow slate]

and ?? umm just i iwll paste the pip cache purge?

like this check?

@ornate elk

[ornate elk]

Yep

[shadow slate]

ok it said files removed 8

relaunching

umm its still in process but it showed this warning

[ornate elk]

Thats okay, thats because you used python 3.10.6 which is a bit old

It will still work

[shadow slate]

ok

@ornate elkyeah it launched

one more thing i installed animate diff

but its not showing in the upper sectio

@ornate elkhello

[sudden pike]

wait for an answer

he's a human being with a life

[shadow slate]

sorry

actually i am trying this since morning

[ornate elk]

It should be under txt2img

Below the steps

[shadow slate]

oh one more thing the faces

are so weird and its not making video its something ti made weird

@ornate elk

[ornate elk]

I won't explain how AnimateDiff works but maybe you first start with images

AnimateDiff needs special files and settings to work. And on a 3050 it won't run that good

[shadow slate]

NO IAM using defrom

[ornate elk]

Ah okay

Can't help with deforum sry, I never used it

But for face fix you can install the Adetailer extension

[shadow slate]

OK ANIMATE DIFF HELP ME NA I I CAN TRY

[ornate elk]

Sry dont have the time. I'm at work

[dry hawk]

is there any way to get generation settings not from stable diffusion client in browser?

[inner patio]

Telegram bot you mean for example?

[dry hawk]

yep, or somewhere else online

[vocal burrow]

not sure i understand, but settings are stored in picture's metadata. So any exif website / program viewer should work.

[dry hawk]

i couldnt find anyone

[vocal burrow]

eg

[inner patio]

Need someone who can merge me a model with 8 Loras.

I cant to it because I have only CPU work with my SD. (MacBook Pro 2019. 🫠)

And now im Not able to merge this.

So, if we found someone, ill pay them some Coffees!

[fair oxide]

How do I Fix CNet Error Loading Script?
ControlNet preprocessor location: J:\SD Forge\models\ControlNetPreprocessor *** Error loading script: main.py Traceback (most recent call last): File "J:\SD Forge\modules\scripts.py", line 525, in load_scripts script_module = script_loading.load_module(scriptfile.path) File "J:\SD Forge\modules\script_loading.py", line 13, in load_module module_spec.loader.exec_module(module) File "<frozen importlib._bootstrap_external>", line 883, in exec_module File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed File "J:\SD Forge\extensions\openpose-editor\scripts\main.py", line 14, in <module> from basicsr.utils.download_util import load_file_from_url ModuleNotFoundError: No module named 'basicsr'

[ornate elk]

Go into the stable-diffusion-webui folder and click in the explorer bar (not searchbar)
There Type cmd and hit enter and then run these two commands one by one:

venv\scripts\activate

pip install basicsr

Then relaunch.

[fair oxide]

I did already but that doesn't fix it. Just gives me a new error.

[ornate elk]

Which error?

[fair oxide]

I tried it from here https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/15094 but now I get a new error
ControlNet preprocessor location: J:\SD Forge\models\ControlNetPreprocessor *** Error loading script: main.py Traceback (most recent call last): File "J:\SD Forge\modules\scripts.py", line 525, in load_scripts script_module = script_loading.load_module(scriptfile.path) File "J:\SD Forge\modules\script_loading.py", line 13, in load_module module_spec.loader.exec_module(module) File "<frozen importlib._bootstrap_external>", line 883, in exec_module File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed File "J:\SD Forge\extensions\sd-webui-txt-img-to-3d-model\scripts\main.py", line 182, in <module> sink: gr.components.IOComponent, AttributeError: module 'gradio.components' has no attribute 'IOComponent'. Did you mean: 'Component'?

Also, I dont think I had ANY errors before I installed Extensions. I removed Openpose edit but still get errors.

Hmm, actually, openpose edit doesnt seem to be removed

even tho I deleted the folder

[ornate elk]

You have to delete these old extensions and then you need to delete the venv folder to get them completely deleted

Because sometimes they load stuff into the venv

The error from above is from a txt to 3d extension

[fair oxide]

alright ill deal with that

Weird. At first it was Openpose in that error, I removed Open pose. But I see it came back, I deleted the Venv folder and restarted.
ControlNet preprocessor location: J:\SD Forge\models\ControlNetPreprocessor
*** Error loading script: main.py
Traceback (most recent call last):
File "J:\SD Forge\modules\scripts.py", line 525, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "J:\SD Forge\modules\script_loading.py", line 13, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "J:\SD Forge\extensions\openpose-editor\scripts\main.py", line 14, in <module>
from basicsr.utils.download_util import load_file_from_url
ModuleNotFoundError: No module named 'basicsr'

Now its the Text to 3D tool.

[ornate elk]

Also make sure your forge is updated

[fair oxide]

How?

[ornate elk]

In the webui-user.bat
Add
Git pull
To the second row/line

Then save and relaunch

[fair oxide]

gotcha

[ornate elk]

Any errors now?

[fair oxide]

I removed the Text to 3D extension. Restarted, error gone. Then I backed up my Installation. Then launched to make sure no errors. No errors. Good, so i removed Venv again just to be sure, and put Git Pull, now its running.

[ornate elk]

Alright good

[fair oxide]

Also, 1 more Question. I would like to specify a Models folder to use. I dont want to keep a seperate Models folder for all my different AI tools (A111 normal, Forge, Reforge)

How can I do that?

outputs too tbh

[ornate elk]

Thats pretty easy to do.
All of these 3 webuis support the command
--models-dir "path to models"

In that main folder you have to name them like in forge or Auto1111

So Lora
Stable-diffusion (for the models)
Controlnet
Etc

[fair oxide]

k

[ornate elk]

Outputs path can be define in the webuis settings under paths

[fair oxide]

Alright, everything loaded mostly normally.
Ill fix this again each time I update NVIDIA GeForce RTX 5080 with CUDA capability sm_120 is not compatible with the current PyTorch installation.

so yeah all good now, thanks

[ornate elk]

No problem 🙂

[humble lily]

RuntimeError: CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

this is what i get when trying to use forge ui, stable diffusion webui, etc

gpu: 5070 ti

[fair oxide]

Did you try doing this? https://www.youtube.com/watch?v=PtGgjdw5koA

[humble lily]

i will see

[fair oxide]

Yeah, I have a 5080 and used to couldn't use some installs of A1111 fresh out the box. Gotta tweak it a bit.

[humble lily]

so i followed the video and it no longer throws out this error but it generates a black image instead every single time. Im using flux1 v2 with ae safetensors

[summer axle]

https://civitai.com/images/84609711
I simply cannot get this to render locally the same.

I am pretty familiar with taking images from civitia and putting them in stable-diffusion webui. I can get them exact most of the time. For this I did PNG info, I have all the LORAs set right. The image just doesn't come out the same with the same seed. Now it's not minor differences. When I generate locally the image is much more zoomed out. I would say the local image is abnout 85% similar. I have poured over the metadata and I just cant figure out what is different. Is there something I am missing on my webui settings that civitai does that I am not aware of?

[fair oxide]

i had to to do that video above to get it to work, not sure why its not working for u

[humble lily]

what model are u using

and it does work as in it doesnt show the error anymore

but generates a black image for some reason

[fair oxide]

Mine is working. Try opening command prompt in the Installations root folder again and do this again
venv\Scripts\activate
then
python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128
after that, close CMD
then run Stable diffusion and close it, and run it again

(mine is working with normal SD models, haven't use anything else other than 1.5 models)

[vestal ledge]

hello, i'm trying to install stable diffusion on a Linux ubuntu 25.04 without cuda. i've found a way to get around the fact that i'm using python 3.13 and i've created the appropriate environment with a Python 3.10.14 version
it starts up fine on Automatic1111 but obviously not everything loads. For example, GFPGAN and codeformer don't start. So face restoration doesn't start. I'm in priory with the latest version of everything. What path should I follow to diagnose what's wrong?

[fair oxide]

XL models are working for me too

[fair oxide]

testing flux soon, downloading crap @humble lily

[humble lily]

alr

[fair oxide]

Flux works too. I made a picture of a Wolf. Just do those steps

[fair oxide]

this

[humble lily]

maybe its because i installed forge using stability matrix?

[fair oxide]

probably

[humble lily]

i will just reinstall everything and without sm

[fair oxide]

I installed Flux by copying theirs using GIT

@humble lily Also 1 more thing you need to know. You will have to redo those steps, every time you delete the Venv folder.
venv\Scripts\activate
then
python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

or else it will break

[azure magnet]

How to fix it?

[humble lily]

just nah

i reinstalled everything from scratch, still nothing

ok it generated

with dpm++ 2M

[fair oxide]

Where it says UI at the top, press Flux
Where it says Vae select a Flux vae, AND in that same spot, it should allow you to select a flux clip thingy. So it should be Clip_I, and t5xxl, and a VAE, in the Vae spot

[humble lily]

whatever this is

[fair oxide]

For Flux Model you have to install and use extra stuff.
Clip_I model, a t5xxl model, and a FluxVAE, also select the Flux UI up top.

[humble lily]

"Where it says Vae select a Flux vae" i believe its the vae? ae.safetensors

and where exactly is the " AND in that same spot, it should allow you to select a flux clip thingy. So it should be Clip_I, and t5xxl, and a VAE, in the Vae spot" cause i dont see it

[fair oxide]

ill show u example, 1sec

much easier with a picture

[humble lily]

i do not have those

the only thing it gives me is ae.safetensors

[fair oxide]

You need to them to be able to Generate Flux Models. Download them from Civitai or Google them.

[humble lily]

and the ae.safetensors? good to keep that?

[fair oxide]

No idea if its the original so I'd delete it if you're not sure

Go here and get these files
https://huggingface.co/lllyasviel/flux_text_encoders/tree/main

[humble lily]

i followed this video to install flux https://www.youtube.com/watch?v=eiopGqEBmPA&t=87s

which did say to download this vae

[fair oxide]

Download Clip_I and the t5xxxl and they go into models\text_encoder

Download whatever VAE's for Flux you like from Civitai and place them in Vae folder.

Then, Have Fun @humble lily

[humble lily]

okay it works, thanks

so it was working all along i just didnt set up flux correctly

[prisma moss]

How to check what cuda version is my forge using?

[ornate elk]

Scroll to the bottom of the Browser when forge is open

[pearl burrow]

Hello, I have a problem installing xformers into stable diffusion webui. Can someone point me in the right direction please?

[ornate elk]

Hey, what's your GPU?

And can you provide a full cmd log?

[pearl burrow]

sure Dual Geforce 1080Ti

in SLI

and python -m xformers.info gives this

[ornate elk]

Which webui do you use?

[pearl burrow]

AUTOMATIC1111 the 1.10.1 version

[ornate elk]

Can you provide a full cmd log when launching auto1111?

[pearl burrow]

how do I do that?

[ornate elk]

Copy and paste

The cmd text

[pearl burrow]

ah just what's in the cmd when I launch?

[ornate elk]

Hmm looks normal except the xformers error.
Have you installed Auto1111 recently?

[pearl burrow]

updated actually

[ornate elk]

Okay then try this:
Open up a cmd and run
Pip cache purge

Then delete the venv folder of Auto1111
Then relaunch the webui-user.bat

It will recreate the venv with the latest torch and xformers version.

If you then still get the same error, then my guess would be that xformers is not usable with SLI.
In that case you can remove --xformers and instead use --opt-sdp-attention
It has the same performance boost as xformers but needs slightly more vram.

I'm off its late. Let me know if it worked.

[fair oxide]

Anyone know how to get XFormers installed and working on SD Forge for Rtx 5000?
version: f2.0.1v1.10.1-previous-668-gd6b1d188  •  python: 3.10.11  •  torch: 2.8.0.dev20250625+cu128
Ive seen xformers==0.0.30.dev1005 Work with 5000 series before but Have no idea how to make it work in forge. Just gives error message saying that version isn't found.

[pearl burrow]

yeah it worked, but now I have xformers 0.0.23 and not the latest version

[sudden pike]

yo. idk why but its somehow really slow and uses lots of VRAM out of nothing

like it takes 10x longer than before

ok nvm i fixed it

[ornate elk]

Thats the latest compatible version for the pytorch version of Auto1111

[pearl burrow]

ok, and there is no way to upgrade?

[ornate elk]

Nope the GTX cards won't benefit from newer torch versions as they require newer cuda versions which GTX cards dont support

[pearl burrow]

ok, thx for the help

[tawny saddle]

so wut do yall think will come of this, kinda seems like their a bit late, like where have they been since 2022

i agree with the 3 to last, but, i feel like, if disney and universal can make money, they will xd, so, wut do yall think

could this spell the end of generative Ai or, will this make things dramically diffirent for the community

[lime wind]

Guys, First time doing ComfyUI here. What is run_nvidia_gpu_fast_fp16_accumulation ???

[winged granite]

Apologies if that's not the right channel

I am getting error: An upstream service tried to fulfill this request, but a required resource was not found.
I tried:

• using the code on the doc and the code in the google colab sandbox
• both using curl or python
• tried new API key, different models: core, sd3
• my billing shows 730+ credits
• API status is green
• not using a VPN
  No clue what is going on

[upbeat oracle]

Hello, what's the problem with API?
https://api.stability.ai/v2beta/stable-image/generate/sd3
ERROR:main:Error Stability AI: 404 - {"errors":["An upstream service tried to fulfill this request, but a required resource was not found."],"id":"b4e3089cc101f2e31dbdca77decd794d","name":"internal_not_found"}

[fading chasm]

Im getting the exact same thing

atleast its not just me

[lime wind]

Hello. I need help with ComfyUI, 1st time using.
Where is the Cross Attention settings? How do I know what Cross Attention its using?

[oak musk]

Stable Assistant
10:11 AM
I apologize, I've encountered an internal error. Please retry your request, and I will do my best to fulfill it. If the issue persists, please let me know so I can address it accordingly.

Anyone else getting this error

[fading chasm]

Im guessing its system wide? Though havent heard if its being looked into

[humble lily]

where can i change my theme to black in SD?

or dark

[ornate elk]

It will be dark if you set windows to dark theme

Or you add --theme dark in the webui-user.bat

[humble lily]

ty

is there a way to save model settings and let them automatically load in each time i select the model? right now i only see the option to save a negative and positive prompt for later use

like a preset for each model that loads in automatically

[ornate elk]

Nope. But you can generate an image with the specific model and settings. Then load the image back into the webui to get them

[junior sparrow]

Hi, I know that you are going to set up comfy UI running environment.
I have hands-on experience build comfy UI running environment on Runpod.
at that time, I configured and tuned each nodes to generate AI image using stable diffusion AI model.
to discuss more, you can DM me.
looking forward to hearing from you.

[lime wind]

This a Bot??? xD

[vocal burrow]

probably

might be

[paper crag]

So i got back from my holiday, tried to open stable diffusion, the thing isn't lauching.
I get these message
Already up to date.
venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.10.1-amd-39-g6d85a7df
Commit hash: 6d85a7dfa4977ae4eece9cbbab4725123833be14
WARNING: you should not skip torch test unless you want CPU to work.
ROCm: agents=['gfx1030']
ROCm: version=5.7, using agent gfx1030
ZLUDA support: experimental
ZLUDA load: path='C:\SD-Zluda\stable-diffusion-webui-directml.zluda' nightly=False
Skipping onnxruntime installation.
version check failed 'commit'
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(

Launching Web UI with arguments: --use-zluda --skip-torch-cuda-test --medvram-sdxl --update-check --skip-ort --theme dark
[-] ADetailer initialized. version: 25.3.0, num models: 30
CivitAI Browser+: Aria2 RPC started
ControlNet preprocessor location: C:\SD-Zluda\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\annotator\downloads
2025-06-29 10:11:26,202 - ControlNet - INFO - ControlNet v1.1.455
Forge: False, reForge: False
Extension: Extra Schedulers: adding new schedulers
Loading weights [a810e710a2] from C:\SD-Zluda\stable-diffusion-webui-directml\models\Stable-diffusion\waiNSFWIllustrious_v130_1512948.safetensors
Creating model from config: C:\SD-Zluda\stable-diffusion-webui-directml\repositories\generative-models\configs\inference\sd_xl_base.yaml
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\huggingface_hub\file_download.py:797: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Applying attention optimization: Doggettx... done.
Press any key to continue . . .

Any suggestions?

[ornate elk]

What's your GPU?

And remove --skip-torch-cuda-test from the webui-user.bat

Then delete the venv folder and relaunch

[paper crag]

AMD Radeon RX 6800

[ornate elk]

You should also consider upgrading to hip SDK 6.2

Much better for the 6000 series

[paper crag]

I'll think about it.
Back to the matter at hand, i think there's an issue when i relauch after deleting the venv

[ornate elk]

Can you relaunch again without deleting the venv?

[paper crag]

Already up to date.
venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.10.1-amd-39-g6d85a7df
Commit hash: 6d85a7dfa4977ae4eece9cbbab4725123833be14
ROCm: agents=['gfx1030']
ROCm: version=5.7, using agent gfx1030
ZLUDA support: experimental
ZLUDA load: path='C:\SD-Zluda\stable-diffusion-webui-directml.zluda' nightly=False
Installing requirements
Skipping onnxruntime installation.
version check failed 'commit'
W0629 11:42:12.144000 2516 venv\Lib\site-packages\torch\distributed\elastic\multiprocessing\redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort --theme dark

[-] ADetailer initialized. version: 25.3.0, num models: 30
*** Error loading script: main.py
Traceback (most recent call last):
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\scripts.py", line 515, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\script_loading.py", line 13, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "C:\SD-Zluda\stable-diffusion-webui-directml\extensions\openpose-editor\scripts\main.py", line 14, in <module>
from basicsr.utils.download_util import load_file_from_url
ModuleNotFoundError: No module named 'basicsr'

---

CivitAI Browser+: Aria2 RPC started
ControlNet preprocessor location: C:\SD-Zluda\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\annotator\downloads
2025-06-29 11:42:19,656 - ControlNet - INFO - ControlNet v1.1.455
Forge: False, reForge: False
Extension: Extra Schedulers: adding new schedulers
Loading weights [a810e710a2] from C:\SD-Zluda\stable-diffusion-webui-directml\models\Stable-diffusion\waiNSFWIllustrious_v130_1512948.safetensors
[LyCORIS]-WARNING: LyCORIS legacy extension is now loaded, if you don't expext to see this message, please disable this extension.
Creating model from config: C:\SD-Zluda\stable-diffusion-webui-directml\repositories\generative-models\configs\inference\sd_xl_base.yaml
Applying attention optimization: Doggettx... done.
Press any key to continue . . .

[ornate elk]

Try deleting the .Zluda folder and the venv and relaunch

If that doesnt help you need a complete reinstall. Best is to use my Guides. I updated them. Stuff has changed.
Like python 3.11.9 is now recommended also you should go for hip SDK 6.2

The Guide is on the first link of the pinned messages.

[ornate elk]

Yep try the fresh install

[paper crag]

sigh... figured as much... so, move all of my stuff, then uninstall it, and do it like in your updated guides?

[ornate elk]

Exactly

Save your models and outputs folder also embeddings if you have some

And styles.csv

Uninstall all of hip SDK 5.7 stuff and Python 3.10.

[paper crag]

yeah, ill probably do if it after i try to upgrade to sdk 6.2 like you suggested

[ornate elk]

You need to uninstall the old hip SDK before installing 6.2

[paper crag]

right, was asking grok for the steps, and it told me double check first for compatibiltiy

[ornate elk]

Pls don't ask ai for that stuff xD

Its okay for generic tasks but for the installation keep following the guide

[ornate elk]

also as gfx1030 is supported nativly by hip sdk 6.2 you dont need any custom gfx files

so its much easier for you to upgrade hip sdk

[paper crag]

yeah, i uninstaled the old sdk programs and downloading 6.2

[craggy prism]

https://api.stability.ai/v2beta/stable-image/generate/sd3

404 Not Found

[orchid quail]

Has anyone tried to use Compel with the stabilityai/sd-x2-latent-upscaler model? I can't get the prompt embeds to match up. I tried setting the embedding type and requires_pooled but I cant make it fit and I don't understand enough to go further

[gritty spear]

I was following the instructions for downloading a1111 and when i got to the second run step im getting this WARNING: Connection timed out while downloading.
WARNING: Connection timed out while downloading.
ERROR: Could not install packages due to an OSError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\Users\Aqua\AppData\Local\Temp\pip-unpack-ubhfx4f8\torch-2.1.2+cu121-cp310-cp310-win_amd64.whl'
Consider using the --user option or check the permissions.

Traceback (most recent call last):
File "C:\Users\Aqua\Desktop\sd1111\webui\launch.py", line 48, in <module>
main()
File "C:\Users\Aqua\Desktop\sd1111\webui\launch.py", line 39, in main
prepare_environment()
File "C:\Users\Aqua\Desktop\sd1111\webui\modules\launch_utils.py", line 381, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "C:\Users\Aqua\Desktop\sd1111\webui\modules\launch_utils.py", line 116, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "C:\Users\Aqua\Desktop\sd1111\system\python\python.exe" -m pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url https://download.pytorch.org/whl/cu121
Error code: 1

am i missing something?

[orchid quail]

Might be permissions, you launched your terminal with admin?

I haven't developed on Windows in 5 years so I dunno

[gritty spear]

i didnt maybe thats why?

im gonna try method two and see if that works

[ornate elk]

Hey, what's your GPU?

[gritty spear]

3090?

[ornate elk]

Then checkout the updated install guides. You find them on the first link of the pinned messages here

[orchid quail]

@ornate elk any wisdom for me? 😂

[gritty spear]

thank you

[ornate elk]

Nope I didnt found a way to use it. Maybe it can work in ComfyUI

[orchid quail]

I'm limiting myself to generating with python code for my project, but thanks 😁

[ornate elk]

Ahh okay

Sry then I can't help ^^

[ornate elk]

NP, let me know if you get an error again. But I think it should work without problems

[gritty spear]

Launching Web UI with arguments: --xformers --theme dark
C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\timm\models\layers_init_.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
Traceback (most recent call last):
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\launch.py", line 44, in main
start()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\launch_utils.py", line 465, in start
import webui
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\initialize.py", line 39, in imports
from modules import processing, gradio_extensons, ui # noqa: F401
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 181, in <module>
bootstrap()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python310\lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
ImportError: DLL load failed while importing cv2: The specified module could not be found.

current issue

[paper crag]

status update, managed to install the new version.
Currently installing my extensions.

thanks for the help

[ornate elk]

Perfect no problem

[ornate elk]

Can you relaunch the webui-user.bat

[gritty spear]

when i relaunch

C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui>git pull
Already up to date.
venv "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.10.1
Commit hash: 82a973c04367123ae98bd9abdf80d9eda9b910e2
Launching Web UI with arguments: --xformers --theme dark
C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\timm\models\layers_init_.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
Traceback (most recent call last):
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\launch.py", line 44, in main
start()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\launch_utils.py", line 465, in start
import webui
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\initialize.py", line 39, in imports
from modules import processing, gradio_extensons, ui # noqa: F401
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 181, in <module>
bootstrap()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python310\lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
ImportError: DLL load failed while importing cv2: The specified module could not be found.

[ornate elk]

Okay, update Python to 3.10.11 64bit.
Then delete the venv folder in the stable-diffusion-webui folder.
Then open up a cmd and run
Pip cache purge
Then relaunch the webui-user.bat

[gritty spear]

hmmm

C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui>git pull
Already up to date.
venv "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.10.1
Commit hash: 82a973c04367123ae98bd9abdf80d9eda9b910e2
Launching Web UI with arguments: --xformers --theme dark
C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\timm\models\layers_init_.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
Traceback (most recent call last):
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\launch.py", line 44, in main
start()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\launch_utils.py", line 465, in start
import webui
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\initialize.py", line 39, in imports
from modules import processing, gradio_extensons, ui # noqa: F401
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 181, in <module>
bootstrap()
File "C:\Users\Aqua\Desktop\sd1111\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python310\lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
ImportError: DLL load failed while importing cv2: The specified module could not be found.

same issue it seems

[ornate elk]

Hmm very strange. Any third party antivirus installed?

Also its not recommended to install the webui on Desktop cause of permission errors and OneDrive sync stuff

Also can you open up a cmd and run:
python --version
And then
Where python
Then screenshot pls

[paper crag]

well this is stupid, i installed some extensions and now i found this new page format that i can't make heads or tails of.
Can anyone help me revert back?

[ornate elk]

Delete the extension from the extensions folder.

Then relaunch

[paper crag]

welp lets see if i can figure out which one it is

[ornate elk]

Looks like lobe/kitchen theme

[paper crag]

i didnt install that one

[ornate elk]

Can you screenshot the extensions folder?

[gritty spear]

i did have it on desktop im trying a clean install on the drive directly

[ornate elk]

Sd-webui-ux

And remove Lycoris extension

Its outdated and not needed

And remove loractl, outdated

[paper crag]

and we're back in business

thanks

[ornate elk]

No problem

[gritty spear]

@ornate elk i re installed followed the reccomended settings from the guide i then got an error then closed and re opened it and got this

venv "C:\A1111\stable-diffusion-webui\venv\Scripts\Python.exe"

INCOMPATIBLE PYTHON VERSION

This program is tested with 3.10.6 Python, but you have 3.11.9.
If you encounter an error with "RuntimeError: Couldn't install torch." message,
or any other error regarding unsuccessful package (library) installation,
please downgrade (or upgrade) to the latest version of 3.10 Python
and delete current Python and "venv" folder in WebUI's directory.

You can download 3.10 Python from here: https://www.python.org/downloads/release/python-3106/

Alternatively, use a binary release of WebUI: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre

Use --skip-python-version-check to suppress this warning.

Python 3.11.9 (tags/v3.11.9:de54cf5, Apr 2 2024, 1012) [MSC v.1938 64 bit (AMD64)]
Version: v1.10.1
Commit hash: 82a973c04367123ae98bd9abdf80d9eda9b910e2
Launching Web UI with arguments: --xformers --no-half-vae --mode dark
C:\A1111\stable-diffusion-webui\venv\Lib\site-packages\timm\models\layers_init_.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
Traceback (most recent call last):
File "C:\A1111\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\A1111\stable-diffusion-webui\launch.py", line 44, in main
start()
File "C:\A1111\stable-diffusion-webui\modules\launch_utils.py", line 465, in start
import webui
File "C:\A1111\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "C:\A1111\stable-diffusion-webui\modules\initialize.py", line 39, in imports
from modules import processing, gradio_extensons, ui # noqa: F401
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\A1111\stable-diffusion-webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\A1111\stable-diffusion-webui\venv\Lib\site-packages\cv2_init_.py", line 181, in <module>
bootstrap()
File "C:\A1111\stable-diffusion-webui\venv\Lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python311\Lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ImportError: DLL load failed while importing cv2: The specified module could not be found.

Python 3.11.9

C:\Users\Aqua>where python
C:\Users\Aqua\AppData\Local\Programs\Python\Python311\python.exe
C:\Users\Aqua\AppData\Local\Programs\Python\Python310\python.exe
C:\Users\Aqua\AppData\Local\Microsoft\WindowsApps\python.exe

[ornate elk]

Make sure your GPU driver is updated.
Uninstall any python other than 3.11.9 via system control panel or system settings / apps.
Then open up a cmd as admin and run:
sfc /scannow
After thats done restart (not shutdown) the PC.
Then delete the venv again.
Edit the webui-user.bat and change --mode dark to --theme dark
Save and relaunch the webui-user.bat

And if the webui opens in the browser make sure its whitelisted in any browser adblocker

And if you have any third party antivirus let me know. They often break stuff

And if you still get the same error try installing Forge Webui from my Guides.

[gritty spear]

i guess ill try that next as that last one didnt work it keep saying importing from timm.models.layers is depricated and to import via timm.layers

even if i understood what that meant i dont klnow how to change how it grabs for either

[ornate elk]

Alright hopefully Forge works then. Its better for your GPU anyway as you can their use flux and chroma too if needed

[paper crag]

say, i think my CivitAi Browser + settings aren't right

i try to browse models

they show me models that im not searching for

like lets say i search Naruto uzumaki, it also shows me models that contain uzumaki in the name

and when i try to insert the link to the model in the search bar it does not show that model, instead its showing me all of the checkpoints

[gritty spear]

WARNING:xformers:A matching Triton is not available, some optimizations will not be enabled
Traceback (most recent call last):
File "C:\forge\webui\venv\Lib\site-packages\xformers_init_.py", line 57, in is_triton_available
import triton # noqa
^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'triton'
xformers version: 0.0.27
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
VAE dtype preferences: [torch.bfloat16, torch.float32] -> torch.bfloat16
Installing bitsandbytes==0.45.3
CUDA Using Stream: True
C:\forge\webui\venv\Lib\site-packages\transformers\utils\hub.py:128: FutureWarning: Using TRANSFORMERS_CACHE is deprecated and will be removed in v5 of Transformers. Use HF_HOME instead.
warnings.warn(
Using xformers cross attention
Using xformers attention for VAE
Traceback (most recent call last):
File "C:\forge\webui\launch.py", line 54, in <module>
main()
File "C:\forge\webui\launch.py", line 50, in main
start()
File "C:\forge\webui\modules\launch_utils.py", line 546, in start
import webui
File "C:\forge\webui\webui.py", line 23, in <module>
initialize.imports()
File "C:\forge\webui\modules\initialize.py", line 32, in imports
from modules import processing, gradio_extensions, ui # noqa: F401
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\forge\webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\forge\webui\venv\Lib\site-packages\cv2_init.py", line 181, in <module>
bootstrap()
File "C:\forge\webui\venv\Lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python311\Lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ImportError: DLL load failed while importing cv2: The specified module could not be found.

i think it just hates me at this point

[ornate elk]

Wow wtf. Okay then something is definitly wrong. Probably still python related. And you didnt answered my question about updated GPU driver and if you have a third party antivirus Programm

@vocal burrow hey, do you have an idea on the cv2 error above?
Happens in Auto1111 aswell as in forge fresh installs

[gritty spear]

i did update and i dont have third party

[ornate elk]

Can check tomorrow again. Its late here.

It has to be something with python

[gritty spear]

ok im sorry i have work all day so I'll respond when i can.

[gritty spear]

maybe i need a lower version?

[ornate elk]

No the version was fine. It works with boath 3.10.11 or 3.11.9 64bit.

Important is to not install them from the Microsoft store

[gritty spear]

didnt do that thankfullt

[vocal burrow]

not sure, haven't read everything since the beginning. And I don't see a full log there, but I do see "Python311" in here.

So I'd say it's either :

• incorrect python installed
• incorrect python is set as default or in the .bat
• venv was built with an incorrect python version and hasn't been cleaned.

In anyways, starting with a clean venv should be a good first step. Rename the venv folder to something like venvBACK and relaunch webui-user.bat

[placid pier]

Is someone familiar with forge?

[junior sparrow]

do you need my help to match embedding type and requires_pooled?
if you want, I can give you my help

[lyric vector]

Help when I install ControlNet using the extensions tab or copy my old controlnet folder my Webui is stuck at commit hash if I remove the folder of ControlNet it will run

I had followed the guide for [NVIDIA] Automatic1111 Webui (Stable-Diffusion-Webui)

[livid jetty]

im trying to launch stable diffusion on AMD, i followed a guide and it says it sould just boot up but mine says (in the command prompt) "module 'torch' has no attribute 'dml' any help?

[livid jetty]

hello?

[ornate elk]

Hey, checkout the first link of the pinned messages for the latest AMD Guides.
If you habe a good GPU go for any of the Zluda guides.

[ornate elk]

Hey, can you screenshot where its stuck?
Python 3.11 should work with controlnet normaly

But you can try uninstall python 3.11.9 and then install 3.10.11 64 bit.
Then delete the venv folder and relaunch.

[livid jetty]

what do you reccomend with zluda? automatic, comfy, swarm, forge, reforge, focus, or sd next?

[lyric vector]

It will stay at commit hash after I installed controlnet It will not proceed even after 10mins or so thats the pic i sent awhile ago If I move my controlnet folder away from extensions folder my webui will proceed normally

What I have done now is to get controlnet from extension tab, after downloading it I then close the webui.bat
Delete venv and open webui once again to reinstall venv then I got this message regarding control net the Green text

After this I can use controlnet normally

Right now this is where am at "ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually."

Trying to fix this error that popping when opening the webui

[ornate elk]

What's your GPU?

I would recommend Forge with Zluda or Auto1111

[ornate elk]

You can add --skip-python-version-check to get rid of the warning at the start.
For controlnet insightface you would need to install visual studio build tools

Or install insightface manualy.

Guide to get Insightface working:

• Install VS C++ Build Tools (if you don't need the whole Visual Studio) and select "Desktop Development with C++" under "Workloads -> Desktop & Mobile": https://visualstudio.microsoft.com/visual-cpp-build-tools/
• OR if you don't want to install VS or VS C++ BT - follow these steps:
  https://github.com/Gourieff/sd-webui-reactor-sfw?tab=readme-ov-file#insightfacebuild

[lyric vector]

Got it! After installing it I don't see any errors anymore Thanks!

By the way is there a way to backup all the extensions and settings on webui so that I can just copy it to another pc and will carry my setup?

[livid jetty]

amd radeon RX 7900 XT

[ornate elk]

You can copy the extensions folder over and the ui-config.json

[ornate elk]

Yea then forge with zluda if you also want to be able to use flux and chroma models in addition.
Or Auto1111 or Reforge if you want to use sdxl/Illustrious models mostly.

[livid jetty]

is there anything that can tell me which can do what? (like and example of what each can do?)

'd like to make videos and images with models i found on citvia. also does the xl and pony stuff matter?

[orchid quail]

Yes please!

[ornate elk]

If you want to make videos then you would need Comfyui.
But its more complex to use

For getting started try out Forge with Zluda.
Xl and pony are model types.
Pony and illustrious are based on sdxl. Any webui can use them.

[paper crag]

@ornate elk i think there is an issue with my CivitAi Browser+ settings.
When i browse somethin, it look doesn't search properly. Say i search 'Son Goku' besides showing me models named 'Son goku' it also shows me models that have just 'son' and 'goku' in their name or in the model description. Also if i try to put the url for the model in the search, it won't show me the model and instead show checkpoints models

[prisma moss]

Anyone here managed to get kohya hrfix extension (or integrated) to actually work on forge??

[prisma moss]

@light elm @pure tusk

[ornate elk]

The extension was updated 8 months ago

I'm pretty sure there are some bugs

[paper crag]

i've been using it since the update, and i never had a bug like this

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

Check the cmd logs of the webui maybe something gives a hint what's wrong

And then the last thing could be that the API of Civitai.com changed maybe

[lament cypress]

guys any ideas what can cuase this?

Im trying to run flux on serverless runpod via forge api, but it keeps giving me broken images..
I treid like 10 different samplers, schedulers and still getting these results

its happening only for flux

[
"/runpod-volume/stable-diffusion-webui-forge/models/text_encoder/clip_l.safetensors",
"/runpod-volume/stable-diffusion-webui-forge/models/text_encoder/t5xxl_fp8_e4m3fn.safetensors",
"/runpod-volume/stable-diffusion-webui-forge/models/VAE/ae.safetensors"
]
}

using vae, clip, txt encoder

[fair oxide]

How do I get this XAdapter Extension running in Auto1111???
https://github.com/showlab/X-Adapter/

[gritty spear]

@vocal burrow@ornate elk any luck on my end?

[livid jetty]

i did the steps for forge with zluda, but it gives me this message when i open user.batit give me this error "AssertionError: Torch not compiled with CUDA enabled"

[ornate elk]

Can you show the full cmd log?

[livid jetty]

couldn't post it because it was too long

[ornate elk]

What's your GPU?

It seems like you didnt installed HIP SDK

[lament cypress]

Is there someone using this on runpod serverless?

https://github.com/ashleykleynhans/runpod-worker-forge

[prisma moss]

is it because hes trying to install both zluda and has --cuda stream?

[celest delta]

So i'm trying to use regional prompter but it doesn't seem to work, despite using the ratio and regions, and BREAK, the characters are still mixing together and doesn't follow what i wrote, dunno what i'm doing wrong

[ornate elk]

No its because he doesn't have hip SDK installed or havent restarted the PC after Installing it. So its not recognised.

[ornate elk]

You can show an example in #📝｜prompting-help

[ornate elk]

Can you uninstall any python and reinstall it and make sure you select "add python to path"
Then open up a cmd and run
Pip cache purge
Then delete the venv of Auto1111 or Forge and relaunch

Then send the log again if it fails

[trail walrus]

Hi. I have just started taking interest in ai generation and Stable diffusion. I downloaded SD and tried to generate an image but I got this error. What should I do?

[ornate elk]

Hey, what's your GPU?
And can you post the full cmd log?

[trail walrus]

NVIDIA RTX 5070 TI

I did some searching and apparently the pytorch downloaded from AUTOMATIC111's UI is not compatible with my gpu

[ornate elk]

Yep. Follow the guide on the first link of the pinned messages.
I added steps for the 50series there too

You have to change the pytorch version manually but its easy

[trail walrus]

okay i started doing a fresh install based on your instructions. im supposed to start the webui-user.bat but in the cmd it wont start and says it couldn't find python...

[ornate elk]

Can you show the cmd log?

Also open up a cmd and type
Python --version
And then
Where python

[ornate elk]

Have you checked the "add python to path" when Installing it?

[trail walrus]

i might have missed this part

[ornate elk]

Then run the installer again, click modify, next, and then select "add python to system environment variables"

[trail walrus]

alright done

[trail walrus]

yep it started the download now

[trail walrus]

okay i did everything but now when i try to start webui-user.bat it looks like this

[ornate elk]

is your GPU Driver updated?

[trail walrus]

there is an update available

[ornate elk]

okay, then do that and then run the torch commands again

[trail walrus]

👍

[trail walrus]

okay now SD seems to be working and i'm able to generate images! thank you so much for your help.

[ornate elk]

perfect! no problem 🙂

[lime wind]

Anybody mind helping me setup this Extension called XAdapter? https://github.com/showlab/X-Adapter

[ornate elk]

This only works in ComfyUI and idk how well

[lime wind]

dammit

[tepid crypt]

im using mimicpc, everytime i start a new session, always takes a time to generate the first image for each checkpoint, why that happens? is there any way to make it only load once and keep it saved?

[urban oasis]

this has probably been asked but im really stuck. trying to use ponydiffusion and this is what is generating. ive been looking all ovewr and fixes ive tried havent been working. again really sorry if this is a common problem im just lost.

[gritty spear]

for forge should i be using the run in the forge file or in the webui?

Creating venv in directory C:\forge\webui\venv using python "C:\Users\Aqua\AppData\Local\Programs\Python\Python311\python.exe"
Requirement already satisfied: pip in c:\forge\webui\venv\lib\site-packages (24.0)
Collecting pip
Downloading pip-25.1.1-py3-none-any.whl.metadata (3.6 kB)
Downloading pip-25.1.1-py3-none-any.whl (1.8 MB)
---------------------------------------- 1.8/1.8 MB 6.1 MB/s eta 000
Installing collected packages: pip
Attempting uninstall: pip
Found existing installation: pip 24.0
Uninstalling pip-24.0:
Successfully uninstalled pip-24.0
Successfully installed pip-25.1.1
venv "C:\forge\webui\venv\Scripts\Python.exe"

INCOMPATIBLE PYTHON VERSION

This program is tested with 3.10.6 Python, but you have 3.11.9.
If you encounter an error with "RuntimeError: Couldn't install torch." message,
or any other error regarding unsuccessful package (library) installation,
please downgrade (or upgrade) to the latest version of 3.10 Python
and delete current Python and "venv" folder in WebUI's directory.

You can download 3.10 Python from here: https://www.python.org/downloads/release/python-3106/

Alternatively, use a binary release of WebUI: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre

Use --skip-python-version-check to suppress this warning.

Python 3.11.9 (tags/v3.11.9:de54cf5, Apr 2 2024, 1012) [MSC v.1938 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-669-gdfdcbab6
Commit hash: dfdcbab685e57677014f05a3309b48cc87383167
Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu121
Collecting torch==2.3.1
Downloading https://download.pytorch.org/whl/cu121/torch-2.3.1%2Bcu121-cp311-cp311-win_amd64.whl (2423.5 MB)
------------------------------- -------- 1.9/2.4 GB 31.7 MB/s eta 016
WARNING: Connection timed out while downloading.
ERROR: Could not install packages due to an OSError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\Users\Aqua\AppData\Local\Temp\pip-unpack-he82p16d\torch-2.3.1+cu121-cp311-cp311-win_amd64.whl'
Check the permissions.

Traceback (most recent call last):
File "C:\forge\webui\launch.py", line 54, in <module>
main()
File "C:\forge\webui\launch.py", line 42, in main
prepare_environment()
File "C:\forge\webui\modules\launch_utils.py", line 428, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "C:\forge\webui\modules\launch_utils.py", line 125, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "C:\forge\webui\venv\Scripts\python.exe" -m pip install torch==2.3.1 torchvision==0.18.1 --extra-index-url https://download.pytorch.org/whl/cu121
Error code: 1
Press any key to continue . . .

working on retrying the install because of the timeout

a weird notification poped up after i unstalled torch it asked if i wanted to remove the limiter on it or something to that effect and i clicked it this time maybe this will help?

[gritty spear]

venv "C:\forge\webui\venv\Scripts\Python.exe"

INCOMPATIBLE PYTHON VERSION

This program is tested with 3.10.6 Python, but you have 3.11.9.
If you encounter an error with "RuntimeError: Couldn't install torch." message,
or any other error regarding unsuccessful package (library) installation,
please downgrade (or upgrade) to the latest version of 3.10 Python
and delete current Python and "venv" folder in WebUI's directory.

You can download 3.10 Python from here: https://www.python.org/downloads/release/python-3106/

Alternatively, use a binary release of WebUI: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre

Use --skip-python-version-check to suppress this warning.

Python 3.11.9 (tags/v3.11.9:de54cf5, Apr 2 2024, 1012) [MSC v.1938 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-669-gdfdcbab6
Commit hash: dfdcbab685e57677014f05a3309b48cc87383167
Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu121
Collecting torch==2.3.1
Downloading https://download.pytorch.org/whl/cu121/torch-2.3.1%2Bcu121-cp311-cp311-win_amd64.whl (2423.5 MB)
---------------------------------------- 2.4/2.4 GB 7.1 MB/s eta 000
Collecting torchvision==0.18.1
Downloading https://download.pytorch.org/whl/cu121/torchvision-0.18.1%2Bcu121-cp311-cp311-win_amd64.whl (5.7 MB)
---------------------------------------- 5.7/5.7 MB 7.9 MB/s eta 000
Collecting filelock (from torch==2.3.1)
Downloading filelock-3.18.0-py3-none-any.whl.metadata (2.9 kB)
Collecting typing-extensions>=4.8.0 (from torch==2.3.1)
Downloading typing_extensions-4.14.0-py3-none-any.whl.metadata (3.0 kB)
Collecting sympy (from torch==2.3.1)
Downloading sympy-1.14.0-py3-none-any.whl.metadata (12 kB)
Collecting networkx (from torch==2.3.1)
Downloading networkx-3.5-py3-none-any.whl.metadata (6.3 kB)
Collecting jinja2 (from torch==2.3.1)
Downloading jinja2-3.1.6-py3-none-any.whl.metadata (2.9 kB)
Collecting fsspec (from torch==2.3.1)
Downloading fsspec-2025.5.1-py3-none-any.whl.metadata (11 kB)
Collecting mkl<=2021.4.0,>=2021.1.1 (from torch==2.3.1)
Downloading https://download.pytorch.org/whl/mkl-2021.4.0-py2.py3-none-win_amd64.whl (228.5 MB)
---------------------------------------- 228.5/228.5 MB 7.8 MB/s

eta 000
Collecting numpy (from torchvision==0.18.1)
Downloading numpy-2.3.1-cp311-cp311-win_amd64.whl.metadata (60 kB)
Collecting pillow!=8.3.,>=5.3.0 (from torchvision==0.18.1)
Downloading pillow-11.3.0-cp311-cp311-win_amd64.whl.metadata (9.2 kB)
Collecting intel-openmp==2021. (from mkl<=2021.4.0,>=2021.1.1->torch==2.3.1)
Downloading https://download.pytorch.org/whl/intel_openmp-2021.4.0-py2.py3-none-win_amd64.whl (3.5 MB)
---------------------------------------- 3.5/3.5 MB 8.0 MB/s eta 000
Collecting tbb==2021.* (from mkl<=2021.4.0,>=2021.1.1->torch==2.3.1)
Downloading tbb-2021.13.1-py3-none-win_amd64.whl.metadata (1.1 kB)
Collecting MarkupSafe>=2.0 (from jinja2->torch==2.3.1)
Downloading MarkupSafe-3.0.2-cp311-cp311-win_amd64.whl.metadata (4.1 kB)
Collecting mpmath<1.4,>=1.1.0 (from sympy->torch==2.3.1)
Downloading https://download.pytorch.org/whl/mpmath-1.3.0-py3-none-any.whl (536 kB)
---------------------------------------- 536.2/536.2 kB 5.8 MB/s eta 000
Downloading tbb-2021.13.1-py3-none-win_amd64.whl (286 kB)
Downloading pillow-11.3.0-cp311-cp311-win_amd64.whl (7.0 MB)
---------------------------------------- 7.0/7.0 MB 7.3 MB/s eta 000
Downloading typing_extensions-4.14.0-py3-none-any.whl (43 kB)
Downloading filelock-3.18.0-py3-none-any.whl (16 kB)
Downloading fsspec-2025.5.1-py3-none-any.whl (199 kB)
Downloading jinja2-3.1.6-py3-none-any.whl (134 kB)
Downloading MarkupSafe-3.0.2-cp311-cp311-win_amd64.whl (15 kB)
Downloading networkx-3.5-py3-none-any.whl (2.0 MB)
---------------------------------------- 2.0/2.0 MB 8.7 MB/s eta 000
Downloading numpy-2.3.1-cp311-cp311-win_amd64.whl (13.0 MB)
---------------------------------------- 13.0/13.0 MB 8.2 MB/s eta 000
Downloading sympy-1.14.0-py3-none-any.whl (6.3 MB)
---------------------------------------- 6.3/6.3 MB 8.6 MB/s eta 000
Installing collected packages: tbb, mpmath, intel-openmp, typing-extensions, sympy, pillow, numpy, networkx, mkl, MarkupSafe, fsspec, filelock, jinja2, torch, torchvision
Successfully installed MarkupSafe-3.0.2 filelock-3.18.0 fsspec-2025.5.1 intel-openmp-2021.4.0 jinja2-3.1.6 mkl-2021.4.0 mpmath-1.3.0 networkx-3.5 numpy-2.3.1 pillow-11.3.0 sympy-1.14.0 tbb-2021.13.1 torch-2.3.1+cu121 torchvision-0.18.1+cu121 typing-extensions-4.14.0
Installing clip
Installing open_clip
Installing xformers
Installing requirements
Installing forge_legacy_preprocessor requirement: fvcore
Installing forge_legacy_preprocessor requirement: mediapipe
Installing forge_legacy_preprocessor requirement: onnxruntime
Installing forge_legacy_preprocessor requirement: svglib
Legacy Preprocessor init warning: Unable to install insightface automatically. Please try run pip install insightface manually.
Installing forge_legacy_preprocessor requirement: handrefinerportable
Installing forge_legacy_preprocessor requirement: depth_anything
Installing forge_legacy_preprocessor requirement: depth_anything_v2
Launching Web UI with arguments: --xformers --cuda-malloc --cuda-stream
Using cudaMallocAsync backend.
Total VRAM 24575 MB, total RAM 32717 MB
pytorch version: 2.3.1+cu121
WARNING:xformers:A matching Triton is not available, some optimizations will not be enabled
Traceback (most recent call last):
File "C:\forge\webui\venv\Lib\site-packages\xformers_init_.py", line 57, in _is_triton_available
import triton # noqa
^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'triton'
xformers version: 0.0.27
Set vram state to: NORMAL_VRAM

Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
VAE dtype preferences: [torch.bfloat16, torch.float32] -> torch.bfloat16
Installing bitsandbytes==0.45.3
CUDA Using Stream: True
The cache for model files in Transformers v4.22.0 has been updated. Migrating your old cache. This is a one-time only operation. You can interrupt this and resume the migration later on by calling transformers.utils.move_cache().
0it [00:00, ?it/s]
Using xformers cross attention
Using xformers attention for VAE
Traceback (most recent call last):
File "C:\forge\webui\launch.py", line 54, in <module>
main()
File "C:\forge\webui\launch.py", line 50, in main
start()
File "C:\forge\webui\modules\launch_utils.py", line 546, in start
import webui
File "C:\forge\webui\webui.py", line 23, in <module>
initialize.imports()
File "C:\forge\webui\modules\initialize.py", line 32, in imports
from modules import processing, gradio_extensions, ui # noqa: F401
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\forge\webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\forge\webui\venv\Lib\site-packages\cv2_init_.py", line 181, in <module>
bootstrap()
File "C:\forge\webui\venv\Lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python311\Lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ImportError: DLL load failed while importing cv2: The specified module could not be found.
Press any key to continue . . .

@ornate elk

venv "C:\forge\webui\venv\Scripts\Python.exe"

INCOMPATIBLE PYTHON VERSION

This program is tested with 3.10.6 Python, but you have 3.11.9.
If you encounter an error with "RuntimeError: Couldn't install torch." message,
or any other error regarding unsuccessful package (library) installation,
please downgrade (or upgrade) to the latest version of 3.10 Python
and delete current Python and "venv" folder in WebUI's directory.

You can download 3.10 Python from here: https://www.python.org/downloads/release/python-3106/

Alternatively, use a binary release of WebUI: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre

Use --skip-python-version-check to suppress this warning.

Python 3.11.9 (tags/v3.11.9:de54cf5, Apr 2 2024, 1012) [MSC v.1938 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-669-gdfdcbab6
Commit hash: dfdcbab685e57677014f05a3309b48cc87383167
Installing requirements
Legacy Preprocessor init warning: Unable to install insightface automatically. Please try run pip install insightface manually.
Launching Web UI with arguments: --xformers --cuda-malloc --cuda-stream
Using cudaMallocAsync backend.
Total VRAM 24575 MB, total RAM 32717 MB
pytorch version: 2.3.1+cu121
WARNING:xformers:A matching Triton is not available, some optimizations will not be enabled
Traceback (most recent call last):
File "C:\forge\webui\venv\Lib\site-packages\xformers_init_.py", line 57, in is_triton_available
import triton # noqa
^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'triton'
xformers version: 0.0.27
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
VAE dtype preferences: [torch.bfloat16, torch.float32] -> torch.bfloat16
CUDA Using Stream: True
Using xformers cross attention
Using xformers attention for VAE
Traceback (most recent call last):
File "C:\forge\webui\launch.py", line 54, in <module>
main()
File "C:\forge\webui\launch.py", line 50, in main
start()
File "C:\forge\webui\modules\launch_utils.py", line 546, in start
import webui
File "C:\forge\webui\webui.py", line 23, in <module>
initialize.imports()
File "C:\forge\webui\modules\initialize.py", line 32, in imports
from modules import processing, gradio_extensions, ui # noqa: F401
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\forge\webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\forge\webui\venv\Lib\site-packages\cv2_init.py", line 181, in <module>
bootstrap()
File "C:\forge\webui\venv\Lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Aqua\AppData\Local\Programs\Python\Python311\Lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ImportError: DLL load failed while importing cv2: The specified module could not be found.
Press any key to continue . . .

[ripe lily]

what is the best stable diffusion for me? NVIDIA GeForce GTX 1650 Ti with Max-Q Design (4 GB), Intel(R) UHD Graphics (128 MB)

im on a laptop btw

[livid jetty]

could somone help me out using loras

i used daugeph on sd to generate shantae, my prompt was just a lora for shantae. but i got this generation

im not dissapointed with it, i love it im just confused

[ornate elk]

Forge Webui should work best for that GPU

[ornate elk]

Make sure the lora is compatible with the base model

[ornate elk]

Can you show your txt2img settings?

[ornate elk]

Okay add --skip - python aversions check
To the webui-user.bat.
Then close everything.

Then go into the webui folder.
Inside the webui Folder click into the File Explorer bar (not searchbar).
Type cmd and hit enter.
Then run the following two commands one by one:
venv\Scripts\activate.bat
pip install opencv

Then relaunch forge

[random turtle]

Hey everyone,

I’ve been wrestling with a strange “dark halo” showing up around masked regions whenever I run a k-sampler / Flux-Redux fill pass in ComfyUI. It only appears on flat backgrounds with a lot of gray in it (white, gray, blues) around masks you can see a subtle darker band even though the underlying flat gray should be uniform.

What I’ve already tried:

sRGB→Linear→k-Sampler→Linear→sRGB sandwich around the sampler to do the interpolation in true linear light.

Edge padding (clamp-to-edge) via the ImagePad KJ node, then cropping back down to avoid sampling black borders.

Post-fill color match / color grade on the filled region to re-tone back to the 50% gray background.

Despite all that I still get the same faint dark shift. I’m pretty sure it’s the gamma bias from sampling in sRGB combined with the mask boundary but I can’t nail it down.

Thanks!

[terse hull]

is there a tutorial for installing a1111 on linux?

[vocal burrow]

depends which linux, also I would install forge instead of a111. It hasn't been updated in months, nearly a year by now.

[terse hull]

manjaro so arch based

[vocal burrow]

so tldr :
1/ install pyenv
2/ download python 3.10.18 using pyenv (or whatever but it must be a 3.10, 3.11 might work but I'd stay away from them just to be safe)
2/ git clone the repo
3/ Using pyenv, set the local python to 3.10.18 in whatever folder you git cloned the repository
4/ run webui.sh and wait.

if you need more precisions let me know

cachyos <=> arch based

[gritty spear]

(venv) C:\forge\webui>pip install opencv
ERROR: Could not find a version that satisfies the requirement opencv (from versions: none)
ERROR: No matching distribution found for opencv

would you like to see my webui text?

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --cuda-malloc --cuda-stream --skip - python aversions check

Then go into the webui folder.
Inside the webui Folder click into the File Explorer bar (not searchbar).
Type cmd and hit enter.
Then run the following two commands one by one:
venv\Scripts\activate.bat
pip install opencv

Then relaunch forge

@REM Uncomment following code to reference an existing A1111 checkout.
@REM set A1111_HOME=Your A1111 checkout dir
@REM
@REM set VENV_DIR=%A1111_HOME%/venv
@REM set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^
@REM --ckpt-dir %A1111_HOME%/models/Stable-diffusion ^
@REM --hypernetwork-dir %A1111_HOME%/models/hypernetworks ^
@REM --embeddings-dir %A1111_HOME%/embeddings ^
@REM --lora-dir %A1111_HOME%/models/Lora

call webui.bat

[quaint delta]

Hello! I haven't posted here in awhile, but I just got Chroma working in Forge WebUI. There's no channel for this. What existing group can be created for this, or do you think it's necessary? I'm creating a lot of really GOOD images I'd like to share but I'm not sure where to put them. Thanks! For example:

[ornate elk]

Okay uninstall everything from python and install python 3.10.11 64bit again.
Then delete the venv folder.
Then open up a cmd and run
Pip cache purge
Then relaunch forge

[prisma plover]

I havent been able to find a clear answer for this, but is sd possible with a AMD 9070? thanks :)

[ornate elk]

Hey, yes its supported.
You can either use zluda with it or the new experimental TheRock from AMD.

Alle guides can be found here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[prisma plover]

okay well that helps narrow it down

i did try a few of your guides for the zluda ones, but it never detected the gpu?

[ornate elk]

Yea I forgot to add the steps that the 9070 users also need to download custom gfx files because hip SDK 6.2 doesn't support them nativly

But I added them a few mins ago

[prisma plover]

oh lifesaver ty

i did see it needed custom files, but i was just winging it and no idea if theyd be the 'right' ones

[ornate elk]

So if you try to install any webui with Zluda you have to do the Step 5

[ornate elk]

One problem is that the custom gfx files are not optimised for the 90xx cards, so the speed is not the best. But the vram management is much better.
TheRock could be faster but as I dont have a 90 series I couldn't compare

[prisma plover]

im sure its faster than the junky old nvidia card i tried with

[ornate elk]

If you want you can setup boath and let me know which is faster ^^

[prisma plover]

yeah i might try that

ill just be happy to get it to work

and when you mean step 5, you mean this right?

🙏

[ornate elk]

I see I need to update the wording at the top xD it now reads like "older than a 6800 or 9070.."

But I mean if it is specifically a 90 series

[prisma plover]

mhm i sorta understood it

cause if its older or A 9070

[ornate elk]

Okay, would it be better if I write it like:
older than a 6800 or IS a 9070/60 ?

[prisma plover]

that could be good, never know with people misreading

definitely clearer

[ornate elk]

Perfect. Thought so too. I'll always try to write it as clear as possible

[prisma plover]

and awesome ty so much for the help! good timing that you just updated things :)

[ornate elk]

No problem, yea i improved the Guides a lot the last days to make them more compact. And then I saw your message and thought "oh there was something I still need to fix"

[prisma plover]

i bet, i mean i cant even begin to wrap my head around some of this, so amazing you even have the guides

[ornate elk]

Yea for AMD Users it was very frustrating at the beginning.
Also there are a lot of outdated or wrong guides/videos out there. So I made that repo and setup everything, testet it and wrote it down.
When HIP SDK 6.4 drops the 90 series will be supported nativly on windows too.

[prisma plover]

oh absolutely all the videos i tried were so old

[ornate elk]

I'm on a 7900xtx, since zluda its very useful for ai. And now with TheRock we got early access native support for them.

[prisma plover]

gotcha

oh and it said after install its supposed to be an error, but the webui just launched for me, im sure thats fine?

[ornate elk]

Yea thats fine, have to update that part again. The dev of the amd fork fixed stuff the last weeks.

[prisma plover]

👍

[ornate elk]

Updated, thx for pointing that out.

[livid jetty]

im trying to use 3d animation FLUX but i get this error

[ornate elk]

You need the clip file for flux it seems

[livid jetty]

gotcha!

thank you

how do i get one

[ornate elk]

Here:
https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main

Put that into models/text_encoder folder

Go for the 5gb file Q8

You also need clip l for flux

You can find it here:
https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main

[prisma plover]

hi im back! it works, but im getting this after a few images.

[livid jetty]

i put t5-v1_1-xxl-encoder-Q8_0 in the text_encoder file but it still says i dont have a CLIP state dict, also do i do anything wit the Clip l file?

[ornate elk]

Yea that message is normal for forge. Do you notice any slower performance after a few images?

[ornate elk]

Clip l goes also in the folder to t5

Then you have to select them in the VAE/clip dropdown at the top of course

Flux also needs the flux VAE file.

[prisma plover]

Yeah it’s much slower after like 3 or so

[livid jetty]

where do i get that?

[ornate elk]

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

[ornate elk]

Can you show me the settings like resolution etc you used?

[prisma plover]

Sure one sec

even set the size small

[ornate elk]

When using hires fix always set the hires steps to 10.

And then at the bottom enable Never oom for tiled vae only.
When upscaling non realistic images.

[prisma plover]

okay gotcha

hopefully that helps, thanks!

[ornate elk]

Np, it should help. Also let me know if you gonna try TheRock to compare the performance.

[prisma plover]

oh right ill try to give that a shot

if i can figure it out lol

[ornate elk]

If not let me know. Its actually a bit simpler than the Zluda setup

[prisma plover]

huh

okay well ill try to get to that! thanks again!

[livid jetty]

and i put it in the same place as the other 2 files you gave me?

[ornate elk]

It goes into models/vae

[livid jetty]

thank you

[livid jetty]

im sorry to keep bothering you but nopw it says i have no VAE stae Dict

[ornate elk]

You have to select it at the top.

Where the clip files are selected

[livid jetty]

which one?

im really sorry. im very new to stale diffusion

[ornate elk]

At the top, not bottom.

Sry

[livid jetty]

here are my settings and stuff, if i need to change anything lmk

[ornate elk]

Where you selected the clip l and t5, you have to also add the AE. Vae

[livid jetty]

im using renderkid3dJavnMerge_miaonova30 checkpoint, and when i generate somting it turns out colorless (almost) and there are white spots are there any fixes to this?

[ornate elk]

That seems like an VAE issue

What's the models base? 1.5 or sdxl?

@livid jetty Ah I see its XL based as its an Illustrious merge.
Here is the VAE you need:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl.vae.safetensors
Put it into the models/VAE folder and select it at the top.

[visual bane]

I'm using depth map with FLUX to create realistic person images with backgrounds, but the person looks unnaturally pasted onto the scene. The background is fine, but the subject doesn’t blend well or appear grounded. Any suggestions to improve realism?

[exotic pawn]

Hi, is there a way to fix this?

also this
RuntimeError: CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
Time taken: 1.4 sec.

but I think they are related to eachother

[ornate elk]

Hey, check the first link of the pinned messages for the installation guides. There are also the steps for rtx 50 series

[exotic pawn]

Yeah I just noticed those, I was gonna say its fixed now

[tacit maple]

I can't get Comfy to work with my new setup.

I recently replaced my old 3070Ti with a 5070Ti

Afterwards I've installed cuda 12.8

nvcc --version returns:

Copyright (c) 2005-2025 NVIDIA Corporation
Built on Wed_Jan_15_19:38:46_Pacific_Standard_Time_2025
Cuda compilation tools, release 12.8, V12.8.61
Build cuda_12.8.r12.8/compiler.35404655_0```

I've read through this thread on Github:  https://github.com/comfyanonymous/ComfyUI/discussions/6643

I've run the ```pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128``` command

I've also run ```pip3 install -U xformers --index-url https://download.pytorch.org/whl/cu128``` because I got an error about xformers and pytorch versions mismatch. It installed and the error was no more.

But I'm still getting:

```Checkpoint files will always be loaded safely.
Traceback (most recent call last):
  File "C:\AI\ComfyUI\ComfyUI\main.py", line 132, in <module>
    import execution
  File "C:\AI\ComfyUI\ComfyUI\execution.py", line 14, in <module>
    import comfy.model_management
  File "C:\AI\ComfyUI\ComfyUI\comfy\model_management.py", line 221, in <module>
    total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
                                  ^^^^^^^^^^^^^^^^^^
  File "C:\AI\ComfyUI\ComfyUI\comfy\model_management.py", line 172, in get_torch_device
    return torch.device(torch.cuda.current_device())
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\AI\ComfyUI\python_embeded\Lib\site-packages\torch\cuda\__init__.py", line 1026, in current_device
    _lazy_init()
  File "C:\AI\ComfyUI\python_embeded\Lib\site-packages\torch\cuda\__init__.py", line 363, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled```

[tacit maple]

Ended up doing a full fresh re-install after backing up the old version. Took me a moment, but it seems to work fine. I'll keep the old version saved just in case, also I'm really happy with the speed increase of my new GPU.

[timid grail]

extremely realistic DSLR photo of a beautiful young woman, light skin, pale complexion, natural beauty, long slightly wavy black hair with middle part, symmetrical and well-proportioned face, full lips, small nose, large brown eyes, thick dark eyebrows, smooth skin texture, expression: tongue out playfully with relaxed eyes, body visible down to thighs or knees, wearing thin see-through camisole, no bra, clearly posed as: standing by window, body turned sideways, one knee lifted slightly, background setting: clean white-walled room with realistic ambient daylight, sunlight casting directional light across wall and bed, full mirror against wall reflecting room elements, rug and slippers partially cut off by frame, strong outfit visibility, confident and natural posture, realistic soft lighting on body and clothing, ambient shadows, high detail skin and facial features, shallow depth of field, 85mm lens bokeh, ultra sharp focus, high resolution, same facial identity, natural skin texture, subtle facial asymmetry, realistic eye reflections

is there any lora i can use to help the facial structure stay intact in tongue out prompts?

[ornate elk]

You can use the Adetailer extension or inpaint the face manually in img2img

[timid grail]

Adetailer does what? also i am switching to comfy ui right now transitioning to it feels weird

[tired apex]

Need to use fooocus ai on amd helpp?

[jolly skiff]

HELLO

[winter radish]

HELLO

[ornate elk]

Adetailer automates the face Inpainting and enhances the face

[ornate elk]

What's your GPU?

[runic void]

How can I use this model? I'm used to just a putting the safetensor files in the models folder for Automatic1111, but this project has four safetensor files. https://huggingface.co/monadical-labs/minecraft-skin-generator-sdxl

[timid trench]

anyone know how i can fix this issue?

i am not a techie

it works fine on the first time but fails to open whenever i reopen it

pls help i have a deadline

i have an nvidia graphics card 4070

will this help me?

[ornate elk]

Click on the first link of the pinned messages. There are the update guides

[timid trench]

thank you

[ornate elk]

It failed because you installed it on Desktop because you have OneDrive sync active on that

[timid trench]

i use the automatic111 right?

okay this time i will do on cdrive

tysm

[ornate elk]

No problem 🙂

[timid trench]

do i need to do anything special to uninstall or simply delete the folders?

[ornate elk]

Simply delete the folder

[timid trench]

@ornate elk is this the file explorer bar?

if so nothing happens when i click cmd on it

[ornate elk]

no thats the searchbar

file explorer bar is on the left

[timid trench]

thank you it works but, the extension im tryingto install this doesnt work

do i need to delete everything to install python 3.10?

[ornate elk]

which extension?

[ornate elk]

you can try uninstall python 3.11, then install python 3.10.11 64bit and then delete the venv folder and relaunch the wbeui-user.bat

[tribal anvil]

How do I get 3 characters in the same image using character loras? I tried using BREAK but it gives me more than 3 and duplicates 1 or 2 instead of giving me 3

[ornate elk]

hey, you therefore need forge-couple extenion for forge or regional-prompter extension for auto1111 or forge

[tribal anvil]

Im pretty sure I have both how do I use them?

[ornate elk]

which webui do you use?

[tribal anvil]

Forge I believe

[ornate elk]

then go with forge couple, here is an example:

[tribal anvil]

Thanks

@ornate elk forge couple is in extensions right I'm looking but can't find it

[ornate elk]

yes then you have to install it first

under extensions, available, load from

[tribal anvil]

I did all that already

I didn't scroll I used Ctrl f

Does it have a different name?

[ornate elk]

oh its not there

then you have to install it trought "install from url"
https://github.com/Haoming02/sd-forge-couple

[tribal anvil]

Stupid question where can I find the download link in the github

I usually click on latest and it takes me to downloads

[ornate elk]

click on green "code" button

then copy

[tribal anvil]

Ok

Got the code but I still can't find forge couple

I pasted it next to load from

Oh install from url

What's the branch and directory name?

[ornate elk]

not needed

[tribal anvil]

Ok thank you

[ornate elk]

np

[timid trench]

im still having trouble installing the extension, it is the depth anything script

[regal flint]

I reinstalled windows and it took out my python with it. I reinstalled python but stable diffusion cant find when I boot it

[timid trench]

did you specify the path?

[regal flint]

I told it to search for paths durring the install. Idk how to do other stuff

[timid trench]

https://www.youtube.com/watch?v=9QrDn_hRSGs try this

worked for me

[stable drift]

Hello, anybody know why in SD the Civitai Browser + is not working. you can use it down but to view the site and know the trigger words. Is down. I really need help in knowing if it will come back. I am using forge

when click on the it takes you there...but nothing happens

[regal flint]

Didn't work

[regal flint]

Nvm I deleted venv and that fixed it

[ornate elk]

Hmmm I can test later if it still works

[knotty bluff]

Excuse me, where should i text a prompt to generate image?

in this discord channel?

[onyx solar]

You don't. Unless you're paying money.

[tiny garden]

Where do I add the git call mentioned in this error message?

note: the directory used is the same I had on that laptop before

[vocal burrow]

you just run it in a cmd

from anywhere

[livid jetty]

could i put any or all of my stable diffusion folder on a usb?

and how do i make a video like this?

also if i were to put sd on a usb, if i were to go to a friends house, would we be able to use sd?

also also, is there a way to make custom settingss for all of my loras and checkpoints? like not having to change to eular a or sd++ or whatever of a prompt and just clicking a lora to change the dimensions?

[vocal burrow]

It won't work unless they have the exact same version of python that you're using at home and you're mounting the usb stick on the same letter.

[livid jetty]

gotcha\

[vocal burrow]

from the moment you launch a1111/forge for the first time it builds a venv folder with tons of hard coded path in there. So you won't be able to move the folder without everything breaking / having to delete the venv folder and rebuild it.

[livid jetty]

alright scratch the usb thing lol

[vocal burrow]

not really, also I don't see how that would be possible.

lora a : resolution 512x512
lora b : resolution 768x768

what happens if you use lora a and b ?

[livid jetty]

they work fine, i just have some loras that require landscape dimensions and i just get tired of changing back and forth all the time ya know?

[timid grail]

Go to extensions folder n delete “depthmap script” cuz its giving error folder already exists n it cant overwrite

then reinstall the extension or just clone it in path using git clone https://github.com/thygate/stable-diffusion-webui-depthmap-script.git

[graceful girder]

Hello, tell me which is the best version (if there is one) between Zulda and DirectML and why if so?

[deep tusk]

Any progress yet? @graceful girder

[graceful girder]

currently some problems with the ZLUDA installation otherwise no

[abstract umbra]

coool

[graceful girder]

any one can help me with thats ? 💀

[ornate elk]

Any zluda version is better than directml

What's your GPU?

[ornate elk]

Wrong vae probably

[livid jetty]

I don't mean to sound entitled but I'm still waiting on an answer, thank you

[cunning girder]

Is there anything similar to "Seed" in Adetailer? To keep one image, but changing the Adetailer result.

[tawny saddle]

think u may have left this in the deepfryer too long

even has grease kinda drippin in the bg xd

or wut looks like grease

[balmy plover]

ok guys serously stuck if anyone is able to help, just spent 7 days creating the perfect model, comfy ui, yet half of te the manager packages i installed did nt end up working. luckily i put it all in a database. but....

any ideas?

[graceful girder]

it's an AMD 6750 XT

[ornate elk]

Okay, and which webui do you currently use?

And for the image please share your settings. Like which model, resolution etc

[tiny garden]

just tried it, but the command can't be found...

[ornate elk]

That could mean python is not in your system path.
Type in a cmd:
Python --version

[tiny garden]

yup, my laptop lacks Python. just installing it

done, but the git command still cannot be called...

[ornate elk]

Have you checked "add python to path" when Installing it?

[tiny garden]

installed it from the Microsoft store

[ornate elk]

Pls uninstall then again

Its highly recommended to install it from the normal python website

As the Microsoft store version causes issues

[graceful girder]

actually zulda and wait i send the images

[ornate elk]

the vae is not for sdxl or illustrious

so thats why the image was broken

you can download and use this one insetad:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl.vae.safetensors

[stray flare]

HELP IM HAVING PROBLEM IN AUTOMATIC1111

MY PC SPEC
I5 12400 F
16GB RAM
RTX 4060(8GB)

the image rendering time is so f#ucking slow like generating single image taking 12 minutes
help i did everything i could 😭

using base model - epic realism / i also tried cyberrealistic still the same issue
@ornate elk 🙏

i also generated image with the same resolution in comfy ui and in fooocus and it generate in less 30 sec

[vast crest]

hi, i installed the wildcards extension. but it doesn't work for me. they told me i have to activate the extension because it will be deactivated. can you tell me how to do it? because i don't know where to look?

[graceful girder]

ohhhhhh okayyy thanks

[ornate elk]

edit the webui-user.bat
at the line COMMANDLINE_ARGS=
you have to add: --xformers --medvram-sdxl
then save and relaunch

also as you only have 16gb of ram you need to increase your windows pagefile.
https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html
Make sure the pagefile is only enabled for the C drive and disabled for any other drive.
Make sure its set to custom for C at 16000min and 24000max.
Then click OK and reboot the PC.

Then you should have the best performance possible for that gpu and ram

[ornate elk]

check the extensions tab of the webui

[tawny saddle]

circle right circle right, left square triangle up

[misty marten]

Can someone help me? I installed automatic11 with zluda and it's giving me this error

[ornate elk]

hey, can you delete the venv folder and relaunch the webui-user.bat again?

[misty marten]

now its giving this error

[ornate elk]

that means you have to download and replace the gfx files for your gpu

like in the guide on step 5

[misty marten]

I just did this, I believe it worked, its compiling now

[ornate elk]

perfect, compiling is needed

it will work

[prisma moss]

@ornate elk do u know about adetailer individual class prompts?

because im trying [CLASS=name] prompt and it isnt applying it at all on said class, even tho its detected

(since ive got multiple classes in a single model)

[ornate elk]

idk, havent heard about adetailer supporting classes

do you mean with different faces or what?

[prisma moss]

apparently its supported by yolo world models, ill have to check it out

[ornate elk]

cant find anything about class support here:
https://github.com/Bing-su/adetailer

[prisma moss]

https://docs.ultralytics.com/models/yolo-world/#available-models-supported-tasks-and-operating-modes

[ornate elk]

ah okay. but that means nothing for adetailer or the webui itself if not implemented by both

so you cant use classes in prompts for the webui. and only use them in the adetailer prompt field if adetailer implemented that feature (which i doubt)
but its still interesting. havent used yolo world much

[tame sun]

Still waiting for Windows ROCM 6.4.1 support.

https://tenor.com/view/gendo-stare-neon-genesis-evangelion-gendo-gif-9367805630101002790

[misty marten]

@ornate elk Do you have the Vae link for illustrious?

[ornate elk]

here:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl.vae.safetensors

[ornate elk]

heard in Q3 they will be there

[tame sun]

Hopefully.

[ornate elk]

with TheRock native pytoch support would be awesome

[tame sun]

Looks pretty cool, yeah.

[quiet matrix]

i keep getting this error, how do i fix this?

i downgraded numpy like it said but still got it

it was functioning normal until yesterday, now it wont let me start the webui at all

[fluid cove]

I'm looking for file:
"control_v11p_sd15_compact_crowd_poses.pth"
or
"control_v11p_sd15_compact_crowd_poses.safetensors"

needed for this installation:
https://civitai.com/models/132721?modelVersionId=147598
Compact Crowds v1.0

I can't find the file anywhere on the internet...
Does anyone have it, or an alternative?

[ornate elk]

Your using an outdated version of forge

[quiet matrix]

i installed directly from here tho(the recommended one)

am i suppose to use the fastest one now?

[ornate elk]

Now the recommended one is fine. But the version in the log was to old. Make sure you run the update.bat once before starting forge

[quiet matrix]

i did

but i did downgrade the Commit hash, maybe that is the problem

i tried using newer versions but my laptop cant run it, requires more performance

that's the reason for me downgrading it and ive been using that version for a the past years

[livid jetty]

help 😭

[livid jetty]

sorry to ask this again, so i cant have an automatic prompt put in when i click a checkpoint?

[trim prairie]

RuntimeError: Couldn't install torch.
Command: "C:\Users\Office\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url https://download.pytorch.org/whl/cu121
Error code: 1

[trim prairie]

help me guys...

i already used python 3.10.6

[ornate elk]

Yea then downgrading was the issue.
Maybe you should try out ReForge. Its better optimised than new Forge

[ornate elk]

We need more info

Full cmd log helps too

[trim prairie]

hope this clear the view

[ornate elk]

Yep you have the wrong python version installed

Uninstall python 3.13 its not compatible

Then install python 3.10.11 64bit and then delete the venv folder before relaunching the webui-user.bat

[trim prairie]

but i have another apps which use conda 3.13 ? how can i brute install this without uninstalling the python 3.13 ?

[ornate elk]

You can install python 3.10.11 with conda too.
Then edit the webui-user.bat and at the line PYTHON=
you add the path to conda 3.10.11

Btw which app uses 3.13 ?

[trim prairie]

facefusion 3.0

[ornate elk]

FaceFusion needs 3.12

[inner dawn]

Some images, if you copy their metadata from pnginfo, have Override settings. How can I set them myself in Stable Diffusion Forge? By default, there are no Override settings in the interface.

[vocal burrow]

not to my knowledge. there might be an extension out there for that maybe. Or you could just use "style" https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#styles, tldr those are shortcuts to save your current settings

[gleaming steeple]

Hi, @everybody

I have an AI image effect issue now and looking for your hellp.

There are some edited phots of original photos and the effect is similar to sketch.
I want to build workflow to upload original photos and get same effected photos.

How to build this kind of workflow?

These are edited photos

And this is original photo example.

Help me to find solution: which is the best model and how to fine-tuning it.
If you can send edited photo of this photo, it would be great

[prisma moss]

https://github.com/newtextdoc1111/adetailer/tree/dev?tab=readme-ov-file#installation-fork-version

Found the modified version of adetailer which supports classes

Will have to test

[ornate elk]

Oh nice! Let me know how it goes

[gleaming steeple]

Hi, @ornate elk
Can you help me?

[ornate elk]

whats the issue?

[lime river]

need recs on which version of Chroma to use on my setup, is the best place to ask here?

[ornate elk]

Sure

What's your GPU?

[lime river]

Using a laptop 3080ti with 16GB VRAM and 32GB of RAM

[ornate elk]

then go for the Q8 or Q4 gguf variant

[lime river]

Got it, thank you!

[ornate elk]

no problem

[lime river]

Wait the huggingface just shows the safetensor files, which I assume I need, and the latest is chroma unlocked v42, not seeing any variants though?

[ornate elk]

you get the gguf variants which are smaller here:
https://huggingface.co/silveroxides/Chroma-GGUF/tree/main

the normal chroma (17.8gb) would be to large for your vram (16gb)

[lime river]

Would BF16 not work since I have 16GB VRAM? not sure if those numbers correlate

[ornate elk]

yes bf16 is to large for your gpu too

the model always needs to fit into the vram

if not it will be very very slow or wont work at all

[lime river]

Got it, will run Q8 for now then

[prisma moss]

It works bro, this fork is huge

[lime pivot]

hello, guys, i have a question, does monitor matter at all in any way forr stable diffusion ? (would appreciate if anyone knows if it matters on blender, hunyuan, photoshop, runway too !) Resolution: FHD (1920×1080)

Refresh Rate: 180Hz

[ornate elk]

Doesnt matter. In fact a 1080p is good as it doesnt requires as much of the GPU as a 4k would.

[lime pivot]

thanks as always !

[gleaming steeple]

Hi, @ornate elk
Can you check this message?

[ornate elk]

Thats not a technical question and belongs more into #📝｜prompting-help

[stray flare]

can i run flux in stable diffusion ( gpu - rtx 4060) ???

[ornate elk]

Yes

But if its a GPU with only 8gb vram then its not worth.

[brittle lantern]

@ornate elk if my fren gets a 9070 XT will he get more of a boost on Linux with ROCm or the same if he uses zluda on windows (idk if ROCm is on windows yet)

[ornate elk]

for now it would be faster on linux because HIP SDK 6.4 is not out for windows yet, but when it is then it will be very similar as AMD also creates TheRock which is native pytorch support for AMD.

[brittle lantern]

Got it do u know what the current speed is for 9070 XT?

I’m also planning to do a 9900x and 9070 XT build

[ornate elk]

nope, in zluda its kinda slow as there are only unoptimised gfx files available for the 90xx series.
With HIP SDK 6.4 there will be no custom gfx files needed due the native support.
With TheRock it could be much faster but got no info on that.
on linux i aldo dont know but definitly faster than on zluda rn.

[ornate elk]

seems like a good build but why the 9900x ?

[brittle lantern]

I did want to go 9900 xt for £350 GBP mainly for the photos editing side and stuff the extra cores may come in handy I’m not sure yet it’s gonna be my first “high end” build I will do to replace my 5600 and 6600 xt build so I’m not sure what’s overkill not underpowered for the price completely yet on high end

[ornate elk]

ah okay yea for editing more cores are always good! the combo on that cpu + gpu is also very nice

if it was purely for gaming then i would always go for the 9800x3d xD

also go for 2x32gb RAM 6000Mhz CL30

[brittle lantern]

I will do that and recording and editing most with AI but I don’t want the CPU to be a issue on editing side since I heard x3D chips are weaker than its normal 9700x or 7700x counterpart depending on what gen x3D ur going for on the productivity side but it’s a lot more stronger for gaming

https://uk.pcpartpicker.com/list/MvgB6Q @ornate elk what I was going for

I do like my micro atx builds

[ornate elk]

nice white build, cool case too

oled gaming monitor is crazy xD

i have to much fear about "burn in" to ever get one as a pc monitor xD

[brittle lantern]

That’s usually a issue with older versions of OLEDs

New ones don’t have much of that issue

Once people notice that then price becomes a issue

Which is understandable

But this build isn’t made for my 1080p monitor unless it’s 450hz+ but I rather go 360hz on 1440p with OLED than 1080p 450hz+ TN or IPS

https://uk.pcpartpicker.com/product/f4Hqqs/aoc-agon-ag246fk-241-1920-x-1080-540-hz-monitor-ag246fk

ok price isn’t as bad as i thought but its TN which is ew even as a heavy comp FPS player

Zowie will always be zowie

[vocal burrow]

540hz ... why...

[ornate elk]

im using an Fast IPS Display with 180hz WQHD

the problem with more hz is, you only ever get them in valorant or csgo xD

even with a 7900xtx / 7800x3d i dont get 180fps in cyberpunk maxed out

but for the price the monitor is def great

[brittle lantern]

Don’t have a dedicated reason I’m still going with 360hz OLED but I would be lying that my inner Cs and Valorant didn’t want 540hz even if I fully know I would regret going with it over the OLED

[brittle lantern]

True OLED is still great for single player I know there’s some 144hz+ oleds just none of them are in the UK

Maybe u can take a look if u have any in stock or in country

UK only had 1 240hz OLED and the rest is 360hz+ OLEDs

At 1440p that is

[ornate elk]

ahh okay

120hz would be to less

i think for wqhd, 144hz should be the min and sweetspot for nomal games is 165hz or 180hz. and for fps games yea whatever you can get starting at 240hz xD

[brittle lantern]

I went from zowie 144hz to IPS 240hz and loved it I can’t wait to test 360hz 1440p QD-OLED

With a whole new pc

Time to put my 5600 and 6600 xt build to rest u did well buddy u did well

(But your job isn’t over yet)

[ornate elk]

the 5600 + 6600xt would be a great Steam Machine

[brittle lantern]

Ooooo interesting what would the steam machine be

Never heard of it

If its cool enough I'll try it

[ornate elk]

you know the Steamdeck the Handehld from Valve?

its running SteamOS (based on Arch Linux)

you can get the SteamOS img or other OS Versions like Bazzite or Chimera OS and use them on your old build. So then its like a full Gaming Machine with Console Feeling and UI.

• SteamOS and Bazzite support an additional Desktop Mode you can easily switch to anytime.

[brittle lantern]

Interesting but wouldn't it be weaker than playing games on the newer machine or am I missing som of the point of this 😭

[ornate elk]

you can hook that machine up to your TV as a console to play like couch coop games with friends or get some emulators on it and make it an arcade machine

[brittle lantern]

Ah I see makes sense I do have a 4k smart TV downstairs

But then 6600 xt and 4k might be a issue lmao

[ornate elk]

yea but for emulation it would still be enough. like up to PS3, Xbox360 or Switch 1

and for split screen games

[brittle lantern]

Imagine mw3 OG Coop survival

Split screen

Absolutely loved doing that as a kid on ps3

[ornate elk]

what i always recommend if you have the main rig in an other room than the TV, get yourself Sunshine + Moonlight.
With that combo you can stream your Games fro mthe PC to the TV with nearly 0 Latency.

Moonlight can run on any device you have pluged into the TV but i would recommend using a device with at least 4gb RAM so no FireTV Stick.

[brittle lantern]

The TV i have is a smart TV

Sky Glass that my mum brought

Idk how much ram it does have or anything

But it has a lot of native apps

[ornate elk]

ah okay, then you can check the store for Moonlight, but it depends if sky uses android tv or a self made sky os

but you can also install moonlight on the old PC too so theres that.
What is also a good idea for an older PC is to use it as Game Server

[brittle lantern]

"No, Sky Glass TV does not use Android as its operating system; it operates on Sky OS, which is specifically designed for Sky devices. Sky OS integrates content from various streaming services and provides a unique user experience tailored for Sky's offerings."

"Skys own custom operating system which is the reason there is no App Store so you can't install other apps. Also likely the reason there is no casting or screen mirroring functionality as well"

[ornate elk]

ah well okay, that means you would need an device pluged into it with Moonlight

[livid jetty]

how do i make videos with wan? is there a guide to install it with an amd 7900 XT

[tawny saddle]

imagine tryna make videos with an amd 5600G with 8GB ram, if its even possible, imagine how long that would take to render xd

literal days

using ai anyway

[livid jetty]

i dont get it

[tawny saddle]

wha

if it takes me like 6mins to render a picture in stable diffusion at 512x512 res, imagine tryna make a video xd

would take ages

[livid jetty]

was this directed at me?

[tawny saddle]

nu >->

[livid jetty]

oh my bad

[ornate elk]

You would need to use ComfyUI with Zluda

[vast crest]

Hi

Guys, do you think I should update Python?

and torch??

[ornate elk]

What's your GPU?

[vast crest]

RTX 4070OC 12 GB 🙂

[mossy venture]

Is there any model similiar to Vxp Illustrious ?
i got this error when using Vxp model
RuntimeError: CUDA error: CUBLAS_STATUS_ALLOC_FAILED when calling cublasCreate(handle)

[ornate elk]

Can you show me the full cmd log?
The model doesn't look any different than other Illustrious ones

[ornate elk]

Your webui seems outdated. I can't even tell if its Auto1111 or Forge xD

[mossy venture]

Can i ask a vid tutorial how to install ?

[ornate elk]

Better check the first link of the pinned messages, there are all the Guides.

[south pecan]

Hey! Are there any mods or admins around? I have an idea that could help improve security on the server and keep bots out, nothing intrusive, just a privacy-friendly way to make things safer for the community. 🙂

[lime pivot]

So I downloaded, forge SDXL one-click package https://github.com/camenduru/forge?tab=readme-ov-file

but i got some errors does anyone know how to fix it ?

[ornate elk]

You downloaded the wrong forge version (an outdated fork)

Check the first link of the pinned messages for the updated guides

[lime pivot]

ok thanks !

[prisma moss]

@ornate elk do u know how to log into huggin face with huggingface-cli login

i keep getting that the access token is incorrect

i cant paste it with right click either (which is apparently the way to go)

[ornate elk]

nope sry, didnt used the hf cli before

[prisma moss]

i managed to get in

[ornate elk]

@edgy plaza
Automatic1111 with CPU Mode:
You need to have Git 64bit and Python 3.10.11 64bit installed.

1. Make a new Folder on your Drive (not on Desktop, Downloads, Documents, Programms, Onedrive) and name it Ai for example:
   C:\Ai\
2. You go into the folder you created in this case Ai, then click in the File Explorer bar (not searchbar) and type cmd then press enter.
3. Then you copy and paste this command:
   git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
   Press enter and after its done you can close the cmd.
4. Edit the webui-user.bat (right click, edit), At the line COMMANDLINE_ARGS= You add: --skip-torch-cuda-test --use-cpu all --no-half --precision full --opt-sub-quad-attention
   6.After editing, save and close the bat. Then launch the webui-user.bat
   After the Installation, that can take a while is done, the Webui will autolaunch in your Default Browser.

[edgy plaza]

it is possible to make that what it downloads is not redundant on the D drive like autoamtic itself, not on C for me to download its libraries requirements

[ornate elk]

the files it downloads are temporary needed and you can delete them afterwards by opening up a cmd and type
pip cache purge

[edgy plaza]

i don't understand i mean that to the D drive all its libraries downloaded not to the C drive all in all this is the least of the problems can i talk to you and show you on the screen what is going on i have git and python already downloaded and autoamtic1111

[ornate elk]

yea show it to me maybe

[edgy plaza]

we can go into a voice channel on the server or a private conversation to better understand each other and show each other what i mean as if there was such an option
but that's tomorrow because i'm going to bed if you don't mind because i have i'll say an interesting case with autoamtic1111 and in advance i have chosen a model what i like for ai and it is in the models folder sewn up

[ornate elk]

i dont think i have time tomorrow. but feel free to share some screenshots of the issue or question you got

[trim prairie]

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

[trim prairie]

can help me with this ? cannot run with GPU since i didnt have NVIDIA GPU, can i force it to run with CPU ?

i try to use this first

launch.py: error: unrecognized arguments: --precision-full

[trim prairie]

now i get this error.. should i just delete that on webui-user.bat ?

it works ! but soooo freeeaaaakkkinnnn slooooowwwwwww

[ornate elk]

It needs to be --precision full
Not --precision-full

And yep CPU mode is very very slow

[lime pivot]

@ornate elk hello, bro, thank you for your help yesterday, I have a follow up question:

this is what I did yesterday,

download "https://github.com/lllyasviel/stable-diffusion-webui-forge/releases/download/latest/webui_forge_cu121_torch231.7z"
unzip
click update.bat file
edit in notes "SKIP+VENV=0
run bat

then i closed everything,
then I went in webui_forge_cu121_torch231/webui folder,
then I clicked the File Explorer bar and type cmd "venv\scripts\activate"
result: "The system cannot find the path specified."

do you know how I can fix this ?

[ornate elk]

Hmm okay try set the skip_venv back to 1 and relaunch

I have to edit the guide as forge seems to have a problem with not skipping the venv

If that doesnt work set skip venv back to 0. but then use the webui-user.bat from the webui folder to launch forge

[lime pivot]

I think I may be missing some files. Do I need Git? Also, is my Python version (3.10.6) correct?
Right now, I can launch Forge, but there's an error when generating images.
Here's what happens when I try to launch it from webui-user.bat

[ornate elk]

Yep you need to have git installed and Python 3.10.11 64bit

[lime pivot]

@ornate elk oh ok then i will download that haha. are there any other files? as my pc is new i forgot what other files i downloaded before xd

[ornate elk]

Nope only these two

[lime pivot]

@ornate elk so i downloaded 3.10.11 and git,
and I set skip venv to 0,
and I tried to launch in webui-uiser.bat, but there was an error "ValueError: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost."

[ornate elk]

Can you share the full cmd log?

[tall grove]

Hello!

I need a gaming laptop but also want to be able to train some AI stuff.
I have a budget of 3500€.

Is this a good setup?

Intel Core Ultra 9 275HX
32GB DDR5-6400
Nvidia RTX 5080 175W
1 TB PCIe 5.0 NVMe SSD

Thanks in Advance!

[ornate elk]

I would always recommend not going for a gaming laptop.
Because of these points:

• Stripped down Hardware to fit inside the case. Vendors often build in smaller GPUs like a rtx 4080 mobile has only 12gb of vram than its pc counterpart with 16gb.
• That thing will get crazy hot, no matter how much coolers or fans it has.
• Cause of the bigger hardware that thing will weight so much that its not easy to transport if you often switch the place.
• You pay more and essentially get less value out of your money. (you pay extra because its mobile with a screen)

[tall grove]

Thank you for your answer. Im aware that a desktop computer would be better but I travel a lot and need a laptop to do the ai work and also be able to game.

The RTX 5080 mobile comes with 16 GB VRAM too

[ornate elk]

okay, thats good and the other specs you posted are more than enough to run most ai stuff

vram is the most important for ai and with 16gb its very good for a laptop

personally i would also put in more storage as 1tb is not much

ai models / files are big ^^

[tall grove]

I also would buy a professional gpu but the A5000 is way slower than a RTX 5080 mobile and the A5500 is a little too expensive

ye thats true but I think I am fine with a RTX 5080

[ornate elk]

yea no need for a proffesional one

[tall grove]

once again thanks mate

❤️

[ornate elk]

no problem

also if you can define the specs then i would personally go for an AMD CPU as it needs less Power and the new Ryzen AI chips are very good

but in either way for image or video generation cpu doesnt matter at all

if you plan on using photshop stuff then maybe go for intel as they have usually more cores

[tall grove]

next to image generation i want to use llms and also finetune them

[paper crag]

Question
IF i'm trying to install SD- Koyah
Is there any compatibility issues with AMD?

[ornate elk]

Yes

I made a guide for Kohya with zluda, which is experimental and idk if it works as I couldn't tested it.
But if you want to train a Lora on AMD you can do that by using OneTrainer with Zluda.
Also in my Guide list.

[paper crag]

okay then, lets give onetrainer a shot

[paper crag]

well, managed to installl it.
Now i gotta figure the settings...
You wouldnt happen to know a tutorial how to make illustrious character loras would you?

[ornate elk]

Nope but I linked a simple guide at the bottom

[turbid willow]

Hey guys! Having an error. I havn't used SD in a few weeks, but it worked last time I used it.

Now however, when I try to generate an image, it says "Connection errored out".

I just did an update on SD to see if that would resolve the issue, no go.

Any idea why this is happening?

[ornate elk]

Hey, can you provide a full cmd log?

[turbid willow]

This?

[ornate elk]

Yep

But it looks normal only deforum and openpose extension has some errors

[turbid willow]

Hmm, I wonder why it's not working.

Thanks for checking for me.

[ornate elk]

NP, can't help much with stability matrix.
It just adds an layer of potential issues on tob of webuis xD

[turbid willow]

Maybe I should reinstall the whole thing?

[ornate elk]

Yep that could help

[turbid willow]

Weird, so I figured might as well restart the computer see if that would help, and now when I try to generate an image it is saying "out of memory error"

[ornate elk]

Oh what's your GPU?

[turbid willow]

4070

[ornate elk]

And what you tried as resolution etc?

[turbid willow]

1024x1024

I didn't do anything crazy, not sure why it says I don't have memory

aaaand, now it says connection errored out again lol. Whole thing is probably jsut borked. Rip

I'm not a very tech savy person, I hope I can manage to install it all correctly the second time around. It was so long ago when I first did it that I don't remember all that I did.

[ornate elk]

If you want a clean install without stability matrix checkout the first link of the pinned messages.
There are all the setup guides.

[turbid willow]

I might have to do that, though I don't think I'm smart enough. I can't even get the easy one click install of stability to work atm lol

I just deleted everything and started over, and it is giving me a new error

Starting to think it was a miracle I ever manage to get it to work in the past XD

Wow.. I completely deleted everything and reinstalled stability and the web ui, got it running finally, and when I try to generate an image, it says "Connection errored out" AGAIN lol.. why me!? XD

[ornate elk]

rip xD

[ornate elk]

let me know if you try the manual method

i can help with that

[turbid willow]

Does the manual method have a flashy user friendly window to enter prompts into and stuff or is all command line stuff?

[ornate elk]

its the same ui as forge

just not with stabiliy matrix as launcher

[turbid willow]

Gotcha

And it's not too hard to do? >_>

[ornate elk]

nope ^^

[turbid willow]

I might try tomorrow, when I'm fresh and not filled with bitterness XD

Hmm, it does look easy, almost too easy? I feel like I must be missing something.

This is all I need to do?

Well, beside installing models and stuff.

[ornate elk]

Yes thats all

[turbid willow]

Nice, I can handle that i bet

[ornate elk]

But these instructions are for Automatic1111.
You used Forge before

But forge is even easier to setup xD

[turbid willow]

Oh, so i had the wrong list. Gotcha, I need forge. Thanks for the heads up

[ornate elk]

No problem!

[dusty mortar]

Hey may i ask help for smth ? i'm knew to forgeui and i don't know how to fix faces etc, i'm just getting weird results lmao

[ornate elk]

Hey, what results do you get?

Facefix is easy but you have to install an extension for it.
In Forges webui, click on the extensions tab.
Click on Available,
then on "Load from"
In the list search for Adetailer
Then click install. Then wait a few seconds. Then relaunch Forge.
Then you have the Adetailer in txt2img and img2img.
Just enable it and leave its settings on default for the start.
It will fix any face.

[dusty mortar]

thanks for this, i tried but i got this XD

[ornate elk]

Okay xD

Show your settings pls

[dusty mortar]

i generated and send the img to img2img

it's in img2img

[ornate elk]

You have to use it in txt2img primarily

In img2img you would have to use the same prompt and resolution AS the input image to use adetailer

[dusty mortar]

i got this, the result is really weird lmaoo

[ornate elk]

Wtf

Is this now txt2img tab?

[dusty mortar]

yes yes

[ornate elk]

Can you screenshot the whole site?

[turbid willow]

I've had such bad luck with ad detailer that I stopped using it and just fixed faces manually with inpaint

[dusty mortar]

because to be honest, i also tried generating with comfyui who is super, but it's really hard to understand the whole thing

[ornate elk]

Can you try without using the loras?

[dusty mortar]

i can

i just have to remove the mfrom the prompt ?

it's better i guess

[ornate elk]

Yea much better

[dusty mortar]

the face is not fck up XD

but the lora are the probs so ?

[ornate elk]

Yep they were to strong

[dusty mortar]

too strong ?

[ornate elk]

You have to lower their impact by lowering their number at the end

[dusty mortar]

i changed the strenght here though