https://uberduck.ai/ logo
Join Discord
Powered by
# tacotron-2-support
  • h

    haru0l

    07/20/2022, 10:15 AM
    a fix was alr given irc
  • h

    haru0l

    07/20/2022, 10:15 AM
    use the search function
  • d

    Deleted User

    07/20/2022, 9:44 PM
    Guys?
  • d

    Deleted User

    07/20/2022, 9:44 PM
    I'm starting to think that maybe I'm making a mistake trying to make a voice for Uberduck, since apparently the tutorial is outdated.
  • d

    Deleted User

    07/20/2022, 9:45 PM
    I have an audio file...
  • d

    Deleted User

    07/20/2022, 9:45 PM
    I just want to simply train a model, and throw it onto Uberduck's site...
  • u

    {K EY1} (Kei)

    07/20/2022, 9:50 PM
    Start by splitting the audio file into segments of 10 seconds or less
  • d

    Deleted User

    07/20/2022, 9:52 PM
    So, every sentence into its own WAV file?
  • u

    {K EY1} (Kei)

    07/20/2022, 9:52 PM
    Basically, yeah Sometimes you can fit two sentences in there; that works too!
  • u

    {K EY1} (Kei)

    07/20/2022, 9:54 PM
    (It's actually helpful to have more than one sentence, because it helps the model learn how to make sentence breaks.)
  • d

    Deleted User

    07/20/2022, 9:54 PM
    There are no sentences
  • d

    Deleted User

    07/20/2022, 9:54 PM
    Just several portions
  • u

    {K EY1} (Kei)

    07/20/2022, 9:56 PM
    I can't listen to that since my phone won't download the audio
  • d

    Deleted User

    07/20/2022, 9:57 PM
    So I have 4 .wav 16bit signed PCM files
  • u

    {K EY1} (Kei)

    07/20/2022, 9:58 PM
    Cool
  • d

    Deleted User

    07/20/2022, 9:58 PM
    Each labeled 1.wav 2.wav 3.wav etc
  • u

    {K EY1} (Kei)

    07/20/2022, 10:00 PM
    Now, make a txt file called transcription or list (or whatever else you want to call it) On separate lines, put
    wavs/{number}.wav|{text}
    with {number} being the name of the file and {text} being what's said in the file
  • d

    Deleted User

    07/20/2022, 10:00 PM
    I'll call it list.txt
  • d

    Deleted User

    07/20/2022, 10:00 PM
    So, for example, I write...
  • d

    Deleted User

    07/20/2022, 10:01 PM
    Copy code
    wavs/1.wav|Incoming message.
  • u

    {K EY1} (Kei)

    07/20/2022, 10:01 PM
    Include the |
  • d

    Deleted User

    07/20/2022, 10:01 PM
    Like that?
  • u

    {K EY1} (Kei)

    07/20/2022, 10:01 PM
    There we go
  • u

    {K EY1} (Kei)

    07/20/2022, 10:01 PM
    Yep
  • d

    Deleted User

    07/20/2022, 10:01 PM
    And the same for all 4
  • u

    {K EY1} (Kei)

    07/20/2022, 10:02 PM
    Yeah And then you'll have your dataset
  • d

    Deleted User

    07/20/2022, 10:03 PM
    Like this
  • d

    Deleted User

    07/20/2022, 10:03 PM
    Copy code
    wavs/1.wav|Incoming message.
    wavs/2.wav|Kids need power.
    wavs/3.wav|Operation call to action.
    wavs/4.wav|Brock Lee Coco Lay Superhero Identitities.
  • d

    Deleted User

    07/20/2022, 10:04 PM
    So, I got that 😛
  • d

    Deleted User

    07/20/2022, 10:04 PM
    But how do I get this shoved into a model? 🙂
1...91011...158Latest