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

[Feature]: Add support for SimSwap/Blendswap ONNX #3

Open
LankyPoet opened this issue Oct 9, 2023 · 3 comments
Open

[Feature]: Add support for SimSwap/Blendswap ONNX #3

LankyPoet opened this issue Oct 9, 2023 · 3 comments
Labels
discussion enhancement New feature or request

Comments

@LankyPoet
Copy link

Feature description

I love this extension and am having a lot of fun with it!

Is it possible to add support for the other ONNX files instead of just the inswapper 128 version?
Examples:
https://huggingface.co/netrunner-exe/Insight-Swap-models-onnx/tree/main
There's a SimSwap 512 version I would love to play with.

I tried selecting it in Reactor Force's options, but it throws the error that it was passed 4 arguments, but only expects 3.

Thank you for your consideration, and thank you for making an awesome extension to start with!

@LankyPoet LankyPoet added enhancement New feature or request new labels Oct 9, 2023
@LankyPoet
Copy link
Author

I think what I am actually asking for is a better version of the face swapper algorithm. Example is https://github.com/ThereforeGames/unprompted just incorporated a "facelift" feature with a GPEN quality option and 512 library.

@Gourieff
Copy link
Owner

Gourieff commented Oct 18, 2023

512 SimSwap is much worse than 128 Inswapper
FaceLift uses combination of GFPGAN+GPEN
GPEN has not been developed for a long time, I don't think that it's better than GFPGAN or CodeFormer

The FaceLift algorithm when you use the "Best quality" preset:

  1. Restore the original face of the original image using GFPGAN
  2. Apply the Insightface pipeline
  3. Restore the result with GPEN

So, it slightly changes original face before swapping, then swaps it with Inswapper and then restores with GPEN
I don't think this is a good idea to apply GFPGAN to the original face, because in this case face details could be changed

@Gourieff Gourieff added discussion and removed new labels Oct 18, 2023
@LankyPoet
Copy link
Author

Ahhh OK, thank you. I think I made an assumption that the 512 would be somehow better.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
discussion enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

2 participants