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]: Facial expressions #459

Open
Gomezzzzz opened this issue Jun 5, 2024 · 4 comments
Open

[Feature]: Facial expressions #459

Gomezzzzz opened this issue Jun 5, 2024 · 4 comments
Labels
enhancement New feature or request new

Comments

@Gomezzzzz
Copy link

Gomezzzzz commented Jun 5, 2024

Feature description

ReActor works marvelously, but unfortunately “flattens” the emotions on the faces. For example, “looking frustrated” or “yawning” and so on will give almost neutral facial expressions.
ADetailer helps a bit, but the facial expressions still become more neutral.
Is there any “facial expression toggler” possible in ReActor? :-)
Or would it require creating special models for Reactor, a model for each facial expression, which nobody will do, of course?

I apologize if I wrote obvious nonsense.

@Gomezzzzz Gomezzzzz added enhancement New feature or request new labels Jun 5, 2024
@chuckstarza
Copy link

I'd second that. I've run some experiments with samplers etc but the results are patchy at best. Plus they need to be extremely OTT expressions to have any real effect.

Something to help match the expressions (and things like make-up etc.) on the original face to the swapped would be great

@Pevernow
Copy link

Pevernow commented Jul 6, 2024

+1

@Pevernow
Copy link

Pevernow commented Jul 6, 2024

I have an idea. How about doing partial redrawing under the control of Controlnet after face swapping?

The premise is to have a Controlnet model that can stabilize the face shape, and try to balance the richness of expression and facial similarity by adjusting the weight of this model.

@Pevernow
Copy link

Pevernow commented Jul 6, 2024

Or, I think this kind of face swapping should essentially support expressions. The main reason for not supporting it may be the lack of consideration in this regard during training. Is it possible to retrain an inswapper model in a community way?

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

No branches or pull requests

3 participants