https://uberduck.ai/ logo
Powered by
# machine-learning
  • s
    I usually use splitter.ai
  • j
    Yes it's more consistent but when lalal.ai works it's like hmmmmm
  • j
    True it takes a while to process on a server
  • f
    Training my Mike Wazowski dataset... will release the model when it's done along with the dataset
  • o
    Training my gta vc tommy vercetti dataset,they have 256 voice lines, so it's going to take a long time
  • m
    looks great for an automatically transcribed dataset
  • d
    How does this look?
  • s
  • j
  • p
    Shouldn't matter too much, of course it's handy too see but test the model out while training the model.
  • f
  • f
    Stonks
  • j
    george liquor's lookin good
  • j
    but his synthesis output isn't
  • j
  • j
    Who's this character @JacobLenstar
  • j
    George Liquor
  • j
    a character by John K
  • j
    Oh okay
  • j
    he was in ren and stimpy
  • j
    ?
  • j
    yep
  • j
    oh now he's working okay
  • j
    I heard
  • d
    How'd you fix it?
  • m
    now I wanna see the model itself.
  • w
    Still tryin to fix Movie Woody and... I think I have a feeling that the TS3 and 4 clips might be the issue here. Cause even though Tom sounds the same, there's a hint of old age heard in his current recordings. So I'm gonna try focusing on only material from 1 and 2 and see if that fixes anything which I still doubt but idk
  • w
    that moment when you accidentally replaced the sound files on jim hanks woody 🤦‍♂️
  • u
    @User quacked: number 14. Troll olo jist does not learn to keep his god damn mouth shut.
  • d
    Nice taskbar
1...298299300...1068Latest