https://uberduck.ai/ logo
Join Discord
Powered by
# machine-learning
  • p
    you're using the wrong colab notebook
  • p
  • m
    @User Nice new PFP!
  • p
    😄
  • i
    i was trying to figure it out but suddenly i get this error everytime i try to create a voice
  • h
    i got that before
  • h
    they need to be 22050hz
  • i
    O_O wait you need to tell me that the voices need to be 22050hz?! i didnt know that! D8)
  • h
    is it for all of them?
  • i
    for ace bunnys voice from loonatics unleashed so
  • h
    if thats the case, you can use audacity to make them 22050
  • i
    okie
  • s
    luckily there's ways to batch change the sample rates (like Total Audio MP3 Converter (the name's a bit of a misnomer, it works for most audio types) )
  • d
  • p
    or ffmpeg
  • p
    for f in *.wav; do ffmpeg -i "$f" -ac 1 -acodec pcm_s16le -ar 22050 out/"$f.wav" ; done
  • i
    wait guys is this good or bad?
  • j
    in your list replace .wav with .npy
  • j
    chop chop master onion turned out terrible again
  • j
    "I am your sensei parappa and I will teach you."
  • i
    this is impossible >~<
  • r
    https://f7e93231b772.ngrok.io/ Testing a new model and vocoder, try it out here! The link is temporary and will expire in a few hours.
  • r
    It's a multispeaker model. It has BFDI characters, MLP characters, Vocaloid characters, and a Touhou Project character
  • u
    What repo you use?
  • r
    A repo called CookieTTS, I use my own fork of it: https://github.com/RAYTRAC3R/cookietts
  • r
    I actually wrote a training script for it, I'll post it in #841437191073955920
  • r
    oh wait nvm
  • r
    i dont have access to that channel lmao
  • r
  • u
    Is this tacotron?
1...676869...1068Latest