Best option to train for quality?

[lenibee]

Hey people! 
I was wondering what "Training Method" Would get me the best quality video? By the way, I am using "Photos only" in "data_src" (about 200 of those). However the data_dst video file is from a movie and almost 13-minutes long. So I do have around 15,000 files in under data_src/aligned.
I am a new member to this community so apologies if I get this into wrong section or something.

 

[Groggy4]

Sae or saehd df will give you the best outcome. But 200 src pics against 15,000 dst pics is a big no go. Think of it this way, every angle and expression your dst have. You wanna match with your src. If you are missing a lot of those, it will just get blurry and look like an mutant from futurama. So basically you should look for a much smaller dst suitable for your src. Or dig up vids of your current src, if it exist. I've made df's with src material down to 40 pics. But then we are talking 10-20 sec dst vid with the smallest changes of facial expression and movement.

 

[lenibee]

Sae or saehd df will give you the best outcome. But 200 src pics against 15,000 dst pics is a big no go. Think of it this way, every angle and expression your dst have. You wanna match with your src. If you are missing a lot of those, it will just get blurry and look like an mutant from futurama. So basically you should look for a much smaller dst suitable for your src. Or dig up vids of your current src, if it exist. I've made df's with src material down to 40 pics. But then we are talking 10-20 sec dst vid with the smallest changes of facial expression and movement.

I really appreciate the response ! Thanks