insetGAN for headtransfer (deepfake)
Deepak2405 opened this issue · comments
Can I use insetGAN for headtransfer? That is, take the head and neck part of user1 and seamlessly fit it on the body of user 2?
Yes, you can do that. The easiest way is to generate random sets of heads and bodies and swap. If you wish to do a swap on existing people, you can leverage existing latent space inversion techniques like e4e to find the appropriate face latent. We do not provide code for inversion of the body, but you could implement it analogously to existing optimization-based techniques.
so, do I have to optimize face and body generators everytime (using run_insetGAN) for a single face swap image?
or, is it possible to train generators to generate seamless face and body images?
For the insetGAN method, you have to run the optimization pipeline for every seamless output.