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

    WeegeeFan1

    10/29/2022, 5:11 PM
    I'll make a wav of it in my drive to show
  • g

    Gosmokeless28

    10/29/2022, 5:11 PM
    Only for Tacotron 2
  • w

    WeegeeFan1

    10/29/2022, 5:11 PM
    So they're bad?
  • g

    Gosmokeless28

    10/29/2022, 5:11 PM
    Are you trying to train a Tacotron 2 model?
  • w

    WeegeeFan1

    10/29/2022, 5:11 PM
    most of mine have about 5 to 10 seconds because he was singing
  • w

    WeegeeFan1

    10/29/2022, 5:11 PM
    Yes I am
  • g

    Gosmokeless28

    10/29/2022, 5:11 PM
    Then yeah
  • j

    Justin

    10/29/2022, 5:11 PM
    For Tacotron2 yeah
  • w

    WeegeeFan1

    10/29/2022, 5:11 PM
    Why?
  • w

    WeegeeFan1

    10/29/2022, 5:12 PM
    I'd think it's good since it'll show what silence sounds like
  • g

    Gosmokeless28

    10/29/2022, 5:12 PM
    What?
  • j

    Justin

    10/29/2022, 5:12 PM
    What does silence sound like 😅
  • g

    Gosmokeless28

    10/29/2022, 5:13 PM
    Exactly, lol
  • w

    WeegeeFan1

    10/29/2022, 5:13 PM
    Ok so when you try to do some voices on Uberduck no matter what reference audio you use it comes back with a "matallic" sound in the background. I was assuming it's because the thing doesn't have data on silence
  • w

    WeegeeFan1

    10/29/2022, 5:13 PM
    Is that correct?
  • w

    WeegeeFan1

    10/29/2022, 5:13 PM
    You may have just disproved my thinking
  • g

    Gosmokeless28

    10/29/2022, 5:13 PM
    It's because breath noises are in the voice data
  • g

    Gosmokeless28

    10/29/2022, 5:13 PM
    I like to cut those out before training
  • w

    WeegeeFan1

    10/29/2022, 5:14 PM
    Breath noises aren't in my silence so I think it'll be fine
  • g

    Gosmokeless28

    10/29/2022, 5:14 PM
    Alright
  • w

    WeegeeFan1

    10/29/2022, 5:14 PM
    Now what about echos?
  • w

    WeegeeFan1

    10/29/2022, 5:14 PM
    Some of what I plan to get audio from have a bit of echo
  • w

    WeegeeFan1

    10/29/2022, 5:14 PM
    Being it's him preforming live
  • w

    WeegeeFan1

    10/29/2022, 5:15 PM
    Also should I try to get some normal talking voice data in my dataset along with singing?
  • g

    Gosmokeless28

    10/29/2022, 5:15 PM
    You'll have to use a deveberator for that
  • j

    Justin

    10/29/2022, 5:15 PM
    Echos/reverb...remove them
  • w

    WeegeeFan1

    10/29/2022, 5:15 PM
    My current ones don't have it but I think I'll do that then
  • j

    Justin

    10/29/2022, 5:15 PM
    Use iZotope RX 10
  • j

    Justin

    10/29/2022, 5:15 PM
    It's a paid software tho
  • w

    WeegeeFan1

    10/29/2022, 5:16 PM
    Can Audacity do that at all?
1...282930...74Latest