search

karanvivekbhargava / obamanet

ObamaNet : Photo-realistic lip-sync from audio (Unofficial port)
MIT License
237 stars 71 forks source link

one sample got good result,many samples got bad result. #17

Open liuyingbin123 opened 5 years ago

liuyingbin123 commented 5 years ago

I trained my own data.When i used single sample,lstm would get goo result,but when i used 50 samples,the result was poor.Sorry to bother you.

decajcd commented 5 years ago

I trained my own data.When i used single sample,lstm would get goo result,but when i used 50 samples,the result was poor.Sorry to bother you.

Can you share tools/process.py?

MNCTTY commented 3 years ago

@liuyingbin123 did you create your own person dataset?