https://uberduck.ai/ logo
Join Discord
Powered by
# tacotron-2-support
  • h
    !pip install git+https://github.com/wkentaro/gdown.git
  • h
    lower down
  • w
    What about talknet users? Any updates to the notebooks there?
  • j
    How long do y'all train Tacotron 2 pipeline for?
  • t
    I usually train up to 40 epoch to prevent overtraining.
  • j
    Wow that seems to only take an hour or so. I was doing 6000 lol
  • t
    Which pretrained model do you train on?
  • j
    It says base_model AITCH (1005 speakers)
  • t
    Even 10 epochs are enough
  • t
    If i recall correctly
  • j
    That's incredible
  • t
    6000 is overkill
  • j
    Saved me about a week lol
  • t
    Do you have colab pro?
  • j
    I did for that. If it goes this quick I won't need it now
  • j
    Is multi speaker for different people in the same voice?
  • h
    different people in the same model yes
  • h
    i've had a hard time getting it to work decently though
  • h
    but when it does work it allows for some pretty cool shenanigans
  • h
    speaker mixing my beloved
  • h
    colab pro isn't worth it imo
  • h
    if you wanna pay use vast.ai
  • w
    There needs to be a tutorial to do the different models locally though on windows. That's all I'd need to use my hardware and get better than anything I'd get out of a paid for gpu
  • t
    Speech cuts off early. Is there a way to prevent it synthesis or do i need to do some changes in training?
  • n
    How did I do dataset_source to make it correctly?
  • g
    Paste the link of your dataset
  • n
    It's still giving me warning sign trying to import files. Teach me how to do it with steps.
  • n
    The only thing I did was Import Files still giving me error.
  • n
    I don't know how to solve. dataset_source doesn't work.
  • a
    How can I fix it?
1...767778...158Latest