https://uberduck.ai/ logo
Join DiscordCommunities
Powered by
# machine-learning
  • i
    Russian sounds little weird and crazy
  • i
  • i
  • u
    A
  • i
    a a a?
  • g
    Not sure if it uses A.I but its still cool:

    https://youtu.be/rR03kgXu7PE▾

  • h
    > GPT-3.5 as the brain how would gpt-3.5 be better than gpt-4
  • h
    it uses stammer which isn't ai, it just looks through the spectrogram of all the frames and finds the one with the least difference
  • f
    @hecko have you tried the program
  • h
    stammer? no
  • g
    I would use it but I never really used github and I dont have access to my laptop rn
  • i
    I'm trying to clone a artist voice, with permission of the artist self and the record label. The artist is singing in Swedish: Is there anyone had luck cloning a singing voice in a different language the English? Would love to hear more about the workflow
  • h
    so-vits doesn't care about language luckily
  • h
    it might perform better with english just because there's so much english data but in principle it accepts any language
  • i
    That is true, i did try one other platform, and it came out ok, it was able to pronounce the words, but the voice clone was horrible, there was like 10% similarity. but then again i only used 1 song to create the clone.
  • a
    Hello. I am trying to find a tutorial on how to create a voice model from my own audio. I have all the audio preprocessed, in like 5-10 sec clips, and a csv with all the transcribing done. I just don't know what to do next. Is there a good walkthrough or youtube i can follow to train and create the model from what i have? Thanks 🙂
  • m
    You can read the guide posted in #768215837248716819
  • m
    That's what I use and with a few mishaps, I made my first model.
  • a
    cool! let me check that out. thank you!
  • a
    @Minecraftian47 (make x from y) thanks i'm on my way now. appreciate it

    https://cdn.discordapp.com/attachments/820353681567907901/1100597288121925682/Screenshot_2023-04-25_at_18.40.18.png▾

  • m
    Nice.
  • a
    For reference, this video was super easy to follow. I did have to subscribe to colab because of the size of my model, but it was pretty straight forward :

    https://www.youtube.com/watch?v=d5bHCexR1y4▾

  • g
    It's a shame that the instruction at 5:23 is false.
  • a
    Well you can win them all. I actually skipped that whole part as I already had all that in place
  • g
  • z
    i've used this app!
  • g
    Cool
  • g
    It's a little too complicated for me to use it, though, so I'll just stick to using Open AI's Whisper auto-transcriber.
  • t
    Anthony Fantano talking about the Grimes competition:

    https://www.youtube.com/watch?v=OJ59T2mcZPU▾

  • p
    applesauce
1...105810591060...1068Latest