I am not sure if this will go away with more iterations, it does not seem so. I think it can be solved by use of a mask. On the border between the head and the face a grey band/area appears, I have seen this several times with different swaps after convert. When training it is jot gissible on the prewiev pictures. Is this a common problem? Any solution? I run Dfaker 256pxs, it looks good othervise. All other parameters are default. Extended mask. I have tried Component mask and its maybe a little better.
Want to understand the training process better? Got tips for which model to use and when? This is the place for you
This forum is for discussing tips and understanding the process involved with Training a Faceswap model.
If you have found a bug are having issues with the Training process not working, then you should post in the Training Support forum.
Please mark any answers that fixed your problems so others can find the solutions.
I think I get it, I have sunlight on the left side if the images and electric light on the right sude and this gives the wrong white-balance with the camera. This is not so visible by naked eye so I did not see it first.
I believe this happens with some models. I think reflect padding can fix this in training (I have not tested extensively), otherwise you can look to blend it out at the convert stage.
My word is final