https://uberduck.ai/ logo
Powered by
# machine-learning
  • a
    but the with_ref voice sounds like crap and I don't understand why
  • p
    23 mins of audio, everything default
  • p
    Dont set the epochs higher
  • a
    this is with 1000 wav files and like a few hours of wavs i think? and it still sounds bad
  • p
    Too much data
  • a
    so I should just do 200 epoch and 2000 steps only?
  • p
    Yep
  • p
    And 20 mins of data
  • p
    25
  • a
    for the wavs files, is there a specific length I should try for? E.g. 8 seconds only
  • k
    t would like to install both and try. I might make a relay if it works
  • k
    and ty
  • p
    2-15 seconds
  • k
    very much
  • p
    Okiii!! @Justin can help you!
  • k
    I can't actually speak
  • p
    No problem
  • p
    Yeah. I'm sorry for that, I've met someone unable to speak...
  • k
    so I use cereproc currently
  • k
    ai
  • k
    with grid
  • k
    grid does alot
  • p
    Yeah. Atleast technology does something
  • k
    ya!
  • p
    Talknet or tacotron2 only really is a text to speech engine
  • k
    I have duck voices on my twitch
  • k
    at least
  • p
    There is uberduck tacotron2, but that's very new and not a lot of models are on it, it supports "emotions" kinda
  • p
    What do you mean with duck voices o-o
  • k
    my chat can use crossbreed priscilla
1...770771772...1068Latest