Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Extension reseting image after inpainting #42

Open
eL1fe opened this issue Nov 8, 2023 · 9 comments
Open

Extension reseting image after inpainting #42

eL1fe opened this issue Nov 8, 2023 · 9 comments
Assignees
Labels
help wanted Extra attention is needed

Comments

@eL1fe
Copy link

eL1fe commented Nov 8, 2023

So I installed the extension, added a prompt to uddetailer with default settings, and started generation. A picture was generated, after that the extension came into play, discovered the face, began to change it, but it was reset by 70% to the default face that was generated initially, and as a result, nothing has changed for me. There are no errors in the console either, there is nothing. What am I doing wrong?

  1. Recognition: face_recognition
  2. Inpainting with my prompt: inpainting
  3. Reseting to defaults: reseting
  4. As result - default picture: done

This is console output: terminal
Help me pls, this extension is so cool, but I can't use it for some reason :c

@eL1fe
Copy link
Author

eL1fe commented Nov 8, 2023

My webui-user.bat looks like:

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--upcast-sampling --medvram --no-hashing  --always-batch-cond-uncond --opt-split-attention --opt-sdp-attention --opt-channelslast  

@eL1fe
Copy link
Author

eL1fe commented Nov 8, 2023

Running on RX 6650 XT, this version of web-ui.

@wkpark
Copy link
Owner

wkpark commented Nov 12, 2023

so, you've used uddetailer + controlnet settings: canny (recently added feature)
you can share your image (contains several uddetailer's settings in it) if you don't mind.
(my personal email address is wkpark@gmail.com)


checklist

  • turn off controlnet settings first -> try again.
  • turn on controlnet settings with other options (like inpainting or sketch)
  • to save your time, you can test your settings with inpainting helper

@wkpark
Copy link
Owner

wkpark commented Nov 14, 2023

After some testing, it appears that the canny controlnet you've selected doesn't always work as intended.

@wkpark
Copy link
Owner

wkpark commented Nov 15, 2023

this is a result with your provided prompt and settings
https://github.com/wkpark/uddetailer/assets/232347/fcd5a9b2-848c-47e9-a6c0-dbfb26f392ef

I can't reproduce but I guess there are some bugs to prevent saving the result correctly for your environment.


One more question. does ADetailer work without this issue? https://github.com/Bing-su/adetailer

@eL1fe
Copy link
Author

eL1fe commented Nov 21, 2023

No, ADetailer generates with the same result as Uddetailer

In this case, apparently there is something wrong with my computer, perhaps it does not have enough VRAM at some point and it resets to the default image or something like that :C

@wkpark wkpark self-assigned this Nov 26, 2023
@wkpark wkpark added the help wanted Extra attention is needed label Dec 6, 2023
@CCpt5
Copy link

CCpt5 commented Dec 9, 2023

Interesting as I haven't tried this yet w/ uddetailer but was about to as I am experiencing this same problem on ADetailer. I just started a thread referencing an older one that was closed: Bing-su/adetailer#438

Hopefully someone can find a solution - odd.

@Bing-su
Copy link

Bing-su commented Dec 10, 2023

img2img의 '인페인팅'모드에서만 나타나는 현상으로, 사용자가 직접 캔버스에 마스크로 지정한 영역 이외를 원본으로 되돌리기 때문으로 보입니다.

@CCpt5
Copy link

CCpt5 commented Dec 12, 2023

A user on the ADetailer repo found a workaround that does work for me using AD:

Bing-su/adetailer#438 (comment)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
help wanted Extra attention is needed
Projects
None yet
Development

No branches or pull requests

4 participants