Faceswap Forum

Hello good people,

I have two videos. One is Charlie Sheen interview and other is my mine. Angles are same, lighting almost same, face expressions are rich. I did more thant 120k iterations and im still getting blurry face in a prieview.
Training model is Original, batch size - 64.
Do i need more information or i just need to do even more training? Can i hope to get clear result?

Thank you for answers!

Original model will only take you so far.

However if the previews appear blurry, that's normally down to quality of data,

Then what training model should i take to get normal results? I though normal training is for normal result (lol).
And when you say "quality of data" you mean data resolution or how much do i have it?
Big THANK YOU for help.

You should check out the guides, they cover a lot of this stuff:

app.php/tag/Guide

torzdf wrote: ↑Sun Mar 29, 2020 10:01 am

Original model will only take you so far.

However if the previews appear blurry, that's normally down to quality of data,

My experience is the same the original model's input/output is too small. For decent result I find 128 input/output is necessary.

Also, you have to do a lot of training. Try about 200k iterations.