https://uberduck.ai/ logo
Powered by
# machine-learning
  • r
    and even the one's that looks very small I guess and different values too.
  • r
    But I hope this one might be good soon.
  • r
    Final stop in the deconder and epoches at the 250 before it finished.
  • m
    hmm. still looks off but it seems to be better?
  • m
    hm.
  • m
    the first part is fine but the other half has issues
  • m
    I have had this weird training with one of my voices that constantly refuses to be decent
  • m
    i havent even trained. i would but im not sure if its physically possible for me to do it
  • m
    Honestly, it aint hard, and you dont need a god pc or anything...
  • f
    I finished compiling voice clips of Mr. Herriman, and now I am currently splitting them up in Audacity.
  • f
    h i p p i t y h o p p i t y
  • m
    ill be needing a dataset however.
  • m
    and i dont exactly have one for the thing im gonna train
  • m
    making a dataset isnt exactly hard, just tedious...
  • m
    thats a issue for me. i dont exactly have the time to do that.
  • r
    For mine, I have labeled like 52 lines of Wylie burp's voices and lines, but one of them were short for no reason at all before I have trained and synthesized them but with bad grammar of mumbling.
  • h
    lower the batch size
  • h
    isnt that gonna be hard LMAO
  • u
    lower than 14?
  • h
    since the wavs are just cv
  • u
    jeeeez how slow is that gonna be
  • h
    yeah
  • h
    try like 10
  • u
    aight
  • h
    i wonder why you're getting oom though
  • h
    are your wav lengths too long?
  • u
    how long is too long?
  • h
    longer than 15secs is too long
  • u
    maybe like one or three
  • u
    out of 128
1...530531532...1068Latest