https://uberduck.ai/ logo
Join Discord
Powered by
# talknet-support
  • u

    {K EY1} (Kei)

    09/07/2022, 12:01 AM
    It's in the pipeline training notebook Just switch the pretrained model to crust if you need
  • p

    postmates!!

    09/07/2022, 12:44 AM
    is it this notebook https://colab.research.google.com/drive/1i0UXcvxXRW1yp9p4cymoZg_ZqsFBFyTB
  • u

    {K EY1} (Kei)

    09/07/2022, 12:46 AM
    No
  • u

    {K EY1} (Kei)

    09/07/2022, 12:46 AM
    It's this https://colab.research.google.com/drive/1jF-Otw2_ssEcus4ISaIZu3QDmtifUvyY
  • u

    {K EY1} (Kei)

    09/07/2022, 12:46 AM
    But i'm just now realizing we're in the talknet channel, and that's tacotron, not talknet
  • u

    {K EY1} (Kei)

    09/07/2022, 12:47 AM
    There's not really any way to train good low-data talknet vbs so go for tacotron If you need to train talknet you can always add more data with results outputted from tacotron
  • p

    postmates!!

    09/07/2022, 12:48 AM
    alr
  • w

    Wru

    09/07/2022, 4:04 PM
    When training a model, can I use one really long wav file (30 mins of talking) or do I have to break it up in to lots of wav files that are one sentence each?
  • w

    Wru

    09/07/2022, 4:07 PM
    also, are 'train_filelist' and 'val_filelist' the same file? I can't find much information about them...
  • p

    PixPrucer

    09/07/2022, 4:30 PM
    You ideally want to have every sentence as its separate wav file
  • p

    PixPrucer

    09/07/2022, 4:30 PM
    The notebooks can't segment by themselves, you have to do it manually
  • w

    Wru

    09/07/2022, 4:31 PM
    cool ok, I saw yt video about doing that in audacity
  • w

    Wru

    09/07/2022, 4:32 PM
    do you know if there's a difference between 'train_filelist' and 'val_filelist'?
  • p

    PixPrucer

    09/07/2022, 4:32 PM
    One file is used for training, one is for validating the AI's performance
  • p

    PixPrucer

    09/07/2022, 4:34 PM
    Most often you use the same list as both, but the secondary one can be customised to include only specific lines (which for ex. Aren't pronounced clearly enough)
  • w

    Wru

    09/07/2022, 4:35 PM
    ok i'll keep them the same content for now
  • d

    Dingusmcbingus

    09/13/2022, 5:09 PM
    i keep getting this error saying wav 1 doesn't exist
  • d

    Dingusmcbingus

    09/13/2022, 5:09 PM
    it does btw
  • d

    Dingusmcbingus

    09/13/2022, 5:10 PM
    nothing weird in the list or val list either
  • w

    Wru

    09/14/2022, 11:12 AM
    does lowering the batch size change the quality of the final model?
  • w

    Wru

    09/14/2022, 11:15 AM
    (on step 9 train spectrogram generator)
  • u

    (Dawn) Will Draw Fictional Women

    09/15/2022, 12:19 AM
  • u

    (Dawn) Will Draw Fictional Women

    09/15/2022, 12:19 AM
    why is this unfunny shit in talknet support
  • m

    Mostafa AL-TAMIMI

    09/15/2022, 6:26 PM
    @samkoelle , delete this disgusting Among Us image, also Hank is posting memes in the wrong channel
  • t

    TheRoyalRuby2000

    09/15/2022, 6:41 PM
    Shhhhhhh! There's an imposter among us.
  • u

    (Dawn) Will Draw Fictional Women

    09/15/2022, 7:02 PM
    i dont wanna do this but instakick for softcore
  • p

    PUMPKINEATER

    09/15/2022, 8:10 PM
    "i dont wanna do this"
  • p

    PUMPKINEATER

    09/15/2022, 8:10 PM
  • u

    (Dawn) Will Draw Fictional Women

    09/15/2022, 9:21 PM
    it was partially funny
  • m

    mepc36

    09/16/2022, 4:35 PM
    What steps are you referencing? Could we get a link to the page?
1...91011...74Latest