Faceswap Forum

Dear faceswap：

When training a pair of identities, the light and shadow of the face of the subject A are more prominent, and the light of the face of the subject B is more average. After the training, the boundary of the face of the subject B will appear very obvious.

But after adjusting the B object to the A object, I trained again. After the training, there was no obvious facial boundary.

In the first case above, is the color tone option not set correctly, or is the face of the B object unsuitable?

Thank you

This is why you want to train from a variety of sources for both A and B. Color augmentation will go some way to fixing this, but it cannot imagine information which just does not exist.

This may be partially fixable in convert with color matching options.