https://uberduck.ai/ logo
Powered by
# machine-learning
  • w
    why is woody just so unfixable? am I just expecting too much?
  • w
    ok so, what do you mean by "audio output"?
  • w
    is it the bitrate of the audio files or what is it?
  • p
    you put a sentence in the text prompt
  • p
    and I want to hear the audio that comes out
  • w
    well, here's one sample of the latest model. it's still not very good.
  • p
    how many epochs?
  • p
    ok I took a listen to your dataset
  • p
    the voice is very varying in prosody
  • p
    you should just get like 50-100 more files to get better results
  • p
  • w
    200. like I usually do for all my models.
  • p
    ok, then you should really get more files
  • w
    that's what I've been trying to do, but again, no matter what I throw at it, the results are always the same.
  • w
    normal at first then strokes
  • w
    also for this, I already know about it. But I'm trying to go for a Tom Hanks only model.
  • w
    plus I just submitted the jim hanks version not too long ago.
  • j

    https://www.youtube.com/watch?v=RwW0Tl8R0Zc▾

  • j
    have you used this?
  • w
    yes.
  • j
    oh
  • j
    maybe just look for more tom hanks in general
  • j
    for shrek I used other sources where mike myers does his scottish accent
  • w
    I could try that.
  • j
    tom hanks interview
  • w
    I mean it would just turn a standalone tom hanks voice if I was to use it, but if it helps I'll give those a shot.
  • p
    oh ok
  • p
    as I said I don't know much about US tv and movies
  • s
    Jim (Tom's brother) does the vidja games, Tom does the movies
  • w
    man, if only I could get a clean rip of the 5.1 tracks from all 4 movies, I would've had a much easier time compiling all the data
1...305306307...1068Latest