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

    WeegeeFan1

    11/02/2022, 5:55 PM
    What would it most likely be?
  • w

    WeegeeFan1

    11/02/2022, 5:55 PM
    (Wait for message to send
  • w

    WeegeeFan1

    11/02/2022, 5:55 PM
    The good sounding clips are just the model regurgitating what I gave it but any time I give it something original to sing it sounds terrible, as shown with "Put Your Head On My Shoulder"
  • u

    {K EY1} (Kei)

    11/02/2022, 6:08 PM
    It's probably just not enough data
  • w

    WeegeeFan1

    11/02/2022, 6:09 PM
    I'm training it 2,000 steps and 500 epochs
  • w

    WeegeeFan1

    11/02/2022, 6:09 PM
    Just thought I say that if that's bad
  • u

    {K EY1} (Kei)

    11/02/2022, 7:54 PM
    It's fine
  • w

    WeegeeFan1

    11/02/2022, 9:02 PM
    Ok I'll add more data
  • b

    Big Lez

    11/03/2022, 8:50 AM
    I deleted it for the screen shot... 😅
  • w

    WeegeeFan1

    11/03/2022, 5:41 PM
    Even after adding another 6 minutes of audio (for a total of 14 minutes) it still can't do anything but regurgitate what I taught it
  • w

    WeegeeFan1

    11/03/2022, 5:42 PM
    That's the test. Guess what is regurgitated and what's new
  • u

    (Dawn) Will Draw Fictional Women

    11/03/2022, 5:54 PM
    talknet just generally sucks
  • u

    (Dawn) Will Draw Fictional Women

    11/03/2022, 5:54 PM
    especially with singing data with held notes
  • a

    a nickname?

    11/03/2022, 5:54 PM
    lol
  • u

    {K EY1} (Kei)

    11/03/2022, 7:08 PM
    @WeegeeFan1 curious, did you feed it singing data? If so, you need to get way more data Try 30 minutes next
  • t

    TavvyProds1997

    11/03/2022, 9:02 PM
    Pig scrotes even!
  • w

    WeegeeFan1

    11/03/2022, 10:29 PM
    I would think giving it Data of him singing should help. The thing us that's most of what I made into the model by accident
  • w

    WeegeeFan1

    11/03/2022, 10:31 PM
    That's exactly what I did. The only thing I gave it of just talking is really just him singing in a natural way.. I'll add some talking from interviewe into the dataset
  • w

    WeegeeFan1

    11/03/2022, 10:32 PM
    It's funny when I try to make him talk fast. He starts to scat singing with mostly "zutzutzutzut" being his only words.
  • w

    WeegeeFan1

    11/03/2022, 10:32 PM
    That'd probably because the fastest pronunciation stuff is scat singing
  • w

    WeegeeFan1

    11/03/2022, 10:33 PM
    I'll add another 20 minutes of interviews data on top of the singing.
  • w

    WeegeeFan1

    11/03/2022, 10:33 PM
    That'll be best, right? Or should I delete the singing?
  • w

    WeegeeFan1

    11/03/2022, 10:43 PM
    The issue with interviews is he stutters a lot

    https://youtu.be/DZXUDUH-3PE▾

    https://youtu.be/G-3JuVFJrY8▾

  • u

    {K EY1} (Kei)

    11/04/2022, 1:01 AM
    You can keep it, but if you keep the singing it won't talk properly without reference audio
  • w

    WeegeeFan1

    11/04/2022, 1:07 AM
    I want him to sing not talk
  • u

    {K EY1} (Kei)

    11/04/2022, 1:14 AM
    Yeah then there's no problem with keeping the singing
  • w

    WeegeeFan1

    11/04/2022, 1:38 AM
    Ok I'll do that
  • w

    WeegeeFan1

    11/04/2022, 1:39 AM
    I also can't find much of him normally talking anyway since he was popular in the 30s and only started singing again in the 80s so there isn't much between the 50s and 70s that I can find (especially the 70s)
  • w

    WeegeeFan1

    11/04/2022, 1:42 AM
    Nver mind just found this it'll be perfect

    https://www.youtube.com/watch?v=E4dysU2sHT4▾

  • w

    WeegeeFan1

    11/04/2022, 1:42 AM
    Am I spamming in here? I'm sorry am i
1...343536...74Latest