Apply Texture LoRA - Qwen Image Edit 2509

Apply wood texture to mug.

Apply Texture Lora - Free for all to use

https://huggingface.co/tarn59/apply\_texture\_qwen\_image\_edit\_2509

Space setup by mulimodalart

https://huggingface.co/spaces/multimodalart/Apply-Texture-Qwen-Image-Edit

https://redd.it/1owdj2u
@rStableDiffusion

Warning! Make sure to NOT store your ConfyUI creations in the ComfyUI folder!

So, like the dumbass I am, I just kept creating new folders inside the Output folder in ComfyUIs folder.

Then today there was a problem starting ComfyUI, it said something about having to install python dependencies (or something similar), but it wouldn't proceed due to there already being a .venv folder. I googled it quickly, and it was suggested to just delete the .venv folder, which I did.

What I didn't know was that it out ComfyUI in some kind of "reset everything to defaults" mode, whereupon ComfyUI permanently deleted everything. Several hundred gigabytes of models gone, as well as everything I've ever created with ComfyUI. The models is one thing, I don't have limited bandwidth, and I have a really fast connection. But I'm quite disappointed in losing every image and every video I've created. It's not a super big deal, its nothing important (which is why I didn't have it backed up), but it just feels so incredibly stupid.

So yeah, either don't use the output folder at all, or make sure to create backups.

https://redd.it/1owiicy
@rStableDiffusion

Qwen Lora requires much higher learning rates than Quen Edit 2509? Any tutorials? Or personal experience? What has been discovered about Qwen Lora training in the last three months?

1e-4 doesn't work well with qwen; even with over 2,000 steps, it seems undertrained. But that same value seems acceptable for qwen. edit 2509v

I have little experience. I don't know if it's my mistake. Batch size = 1. I generally practice with 10 to 30 images.

https://redd.it/1owkr5w
@rStableDiffusion

what gives this away as ai?
https://redd.it/1owr2og
@rStableDiffusion

What model have best prompt adherence?

What I'm interested is build complex scene, like two person fighting pointing guns / sword and etc.

https://redd.it/1owqitn
@rStableDiffusion

OpenSource Face Swapper

can i ask what's the best face swapper currently ? i prefer one with open source, thanks !

https://redd.it/1owolx1
@rStableDiffusion

Slow LoRA Training LoRA on 5090
https://redd.it/1owut58
@rStableDiffusion

Created a little story around my favorite artwork in my town (Qwen Edit + Wan 2.2)
https://www.youtube.com/watch?v=8y1MAaHunEA

https://redd.it/1owsdef
@rStableDiffusion

Eigen-Banana-Qwen-Image-Edit: Fast Image Editing with Qwen-Image-Edit LoRA
https://redd.it/1owykn7
@rStableDiffusion

Testing LipSync - Music Video

https://reddit.com/link/1owz6hu/video/d34qmk5lo81g1/player



https://redd.it/1owz6hu
@rStableDiffusion

Please, how do I make these kind of artstyle?

https://redd.it/1ox2831
@rStableDiffusion

[WIP] Another ComfyUI interface, personal project. I can release to the public if there is enough interest.
https://redd.it/1ox6v1m
@rStableDiffusion

TagGUI Video - Make captions for Wan LoRA training
https://youtu.be/FGNSq01nOT4

https://redd.it/1ox7jyy
@rStableDiffusion

Prevent your images, workflows and model from being deleted during ConfyUI update

This post is related to https://www.reddit.com/r/StableDiffusion/comments/1owiicy/warning\_make\_sure\_to\_not\_store\_your\_confyui/ but Reddit would not let me post this as a comment. So I try a separate post.

As many recommended to create "Symlinks", the actual command for this is "mklink"

https://learn.microsoft.com/de-de/windows-server/administration/windows-commands/mklink

I use a combination of

--output-directory <path\to\output> --user-directory <path\to\user>

and Junctions (they work better for me than Symlinks, also as I used Sysinternals junction.exe for many years).

mklink /j ".\ComfyUI\models" "G:\StableDiffusionModels"

This is my launch command in "run_nvidia_gpu.bat" (portable version):

.\pythonembeded\python.exe -s ComfyUI\main.py --windows-standalone-build --output-directory "D:\StableDiffusion\Image-Outputs\Comfy-Outputs" --user-directory "D:\StableDiffusion\ComfyUIUser"

Which works for this directory structure:

D:\StableDiffusion\
├─ ComfyUIportable0357\
│ ├─ ComfyUI\
│ ├─ pythonembeded\
│ ├─ runnvidiagpu.bat
│ ├─ MKLINKCreateComfyUI-ModelsG-ComfyUIModels+User..ComfyUIUser.bat
├─ ComfyUIUser\
├─ Image-Outputs\
│ ├─ Comfy-Outputs
│ ├─SD-WebUI-Outputs\

The user directory and the image output directories are safely outside the ComfyUI folder.

In addition, I copy my batch file "MKLINK\Create_ComfyUI-Models_G-ComfyUIModels+User_..ComfyUI_User.bat"

to each ComfyUI_portable revision folder (here ComfyUI_portable_0357).

Batch content:

echo Off

if exist .\ComfyUI\models\ (
echo Folder ".\ComfyUI\models" already exists. Cannot create Junction.
echo Check content and move folder to new location or delete it, if "models" folder already exists at new location.
goto Exit
)

if exist .\ComfyUI\user\ (
echo Folder ".\ComfyUI\user" already exists. Cannot create Junction.
echo Check content and move folder to new location or delete it, if "user" folder already exists at the new location.
goto Exit
)


Echo Creating Junctions
mklink /j .\ComfyUI\models G:\StableDiffusionModels
mklink /j .\ComfyUI\user ..\ComfyUIUser

echo Done!

:Exit
echo.
pause

Yes I know, the "user" folder is handled twice, in the start bat and with junction, but better safe than sorry.

However to use this batch file, you first need to relocate the model and user folder to a safe location, they must not extist anymore in ComfyUI folder.

Note that if the ComfyUI update batch files are used, the Junctions will be replaced by standard folders from git. So you need to delete thes folder and run the MKlink batch again.

The "custom\nodes" folder is not included on purpose, to keep this separaters in different Comfy installs.

Hope this helps.

Edit, fixed a formatting issue.

https://redd.it/1ox2uc3
@rStableDiffusion