https://uberduck.ai/ logo
Join Discord
Powered by
# tacotron-2-support
  • t
    oh ok
  • t
    The graph thing is giving me this when using the legacy training notebook when at the bottom of the notebook it is telling me that isn't what a good training looks like so how do I fix it?
  • s
    I Was Able To Toss A 40 Minute Video There
  • u
    you did not need to ping me
  • g
    Train a model with a better dataset
  • t
    So like the amount of wav files?
  • t
    I have 484 wav files
  • g
    How many minutes of voice data do those wavs total in duration?
  • t
    I remember it being around 11-13 minutes
  • g
    That amount is fine. IDK why Tacotron2 isn't generating good-looking graphs when training your model.
  • t
    So should I just retrain the model again
  • g
    What difference would that make?
  • u
    what is the average file length
  • u
    the updated dataset version of edd gould sounds terrible, how to fix?
  • s
    Delete The Model, It Is Not Trained Properly
  • u
    ok
  • t
    okay, now i have around 25 minutes of training
  • t
    is that adequate?
  • g
    You mean 25 minutes of voice data, right?
  • t
    Hey Gosmoke, not sure if you noticed but today I made a voice model of Mailbox from Blue's Clues.
  • t
    Sorry for going off topic BTW, but wanted to let you know about a new model.
  • t
    Do you mean 25 minutes of voice data for your dataset?
  • g
    That's cool
  • t
    oh, yeah
  • t
    i did
  • g
    Then yes, that's a good amount of voice data
  • t
    NICE
  • t
    thanks gosmoke
  • g
    No problem
  • t
    okay new problem solved
1...118119120...158Latest