https://uberduck.ai/ logo
Join Discord
Powered by
# machine-learning
  • t

    TonyTony

    09/05/2022, 10:50 AM
    Colab?
  • t

    TonyTony

    09/05/2022, 10:51 AM
    How much data
  • t

    TonyTony

    09/05/2022, 10:51 AM
    I want him to sound like the real deal
  • m

    matty1406

    09/05/2022, 10:51 AM
    as much as you want, the more the better though
  • t

    TonyTony

    09/05/2022, 10:52 AM
    Pages and pages of transcription? I really want this good
  • t

    TonyTony

    09/05/2022, 10:52 AM
    There's no shortage of voice clips
  • m

    matty1406

    09/05/2022, 10:53 AM
    you can go as high as you want, an hour of data is generally good if you want it to sound really good but more data helps it be even better
  • t

    TonyTony

    09/05/2022, 10:53 AM
    Damn
  • t

    TonyTony

    09/05/2022, 10:53 AM
    An hour
  • t

    TonyTony

    09/05/2022, 10:53 AM
    That's no joke
  • t

    TonyTony

    09/05/2022, 10:53 AM
    Ok
  • t

    TonyTony

    09/05/2022, 10:54 AM
    So I gotta like take audio clips
  • t

    TonyTony

    09/05/2022, 10:54 AM
    But how do I transcribe audio clips like
  • t

    TonyTony

    09/05/2022, 10:54 AM
    Do I just make one big audio clip and then one big notepad document with all the words that are being said?
  • m

    matty1406

    09/05/2022, 10:54 AM
    pretty much, it has to be in a certain way tho
  • m

    matty1406

    09/05/2022, 10:55 AM
    Like this:
    Copy code
    wavs/1.wav|Text goes here.
    wavs/2.wav|Speaking text goes here, with commas for short pauses. And full stops for long pauses.
  • t

    TonyTony

    09/05/2022, 10:56 AM
    So when I upload it is it some guy who's feeding it to the bot?
  • d

    Dyno

    09/05/2022, 10:56 AM
    Read #answers, @TonyTony.
  • t

    TonyTony

    09/05/2022, 10:56 AM
    Dyno
  • m

    matty1406

    09/05/2022, 10:56 AM
    you use the colab notebook which runs code for you and does the whole thing
  • t

    TonyTony

    09/05/2022, 10:57 AM
    But, all I've seen is there's a page where it asks for the files and then you put in some info is that what you're talking about
  • t

    TonyTony

    09/05/2022, 10:57 AM
    This thing
  • m

    matty1406

    09/05/2022, 10:57 AM
    that's what you do after you've made the model, so moderators can approve it
  • m

    matty1406

    09/05/2022, 10:57 AM
    notebooks are in #841437191073955920
  • m

    matty1406

    09/05/2022, 10:58 AM
    there's the training notebook, which you run and it gives you a model in your gdrive, then a synthesis notebook to test your model...
  • t

    TonyTony

    09/05/2022, 10:59 AM
    Geeeez this is tough. But I'll have to do it eventually
  • t

    TonyTony

    09/05/2022, 11:01 AM
    I've been trying to learn computer stuff
  • t

    TonyTony

    09/05/2022, 11:01 AM
    It's hard for me cause I'm old and have no patience
  • t

    TonyTony

    09/05/2022, 11:01 AM
    Staring at the screen at the numbers and tutorials gives me a headache
  • m

    matty1406

    09/05/2022, 11:03 AM
    most of the transcribing is just listening to the data and typing out what you hear, over and over until you finish... more tedious than difficult
1...969970971...1068Latest