https://uberduck.ai/ logo
Powered by
# machine-learning
  • s
    i mean it depends on if it's consistent (see Howard Finkel)
  • s
    @IBob012 made his own hifi model trained on reverb voices and it worked great
  • i
    yeah it doesnt fix the broken aligment reverb usually brings but it makes it sound better
  • s
    Shite
  • s
  • i
    if the reverb isnt big (like circus baby) it could work, but if the reverb is like the smash bros ultimate announcer, it wont be able to say more than 2 words without dying most of the time
  • v
    Hm okay, thanks for the advice guys <3
  • v
    Another question, when transcribing a regular voice, I tend to use “-“ to separate stutters or words that are said in quick succession. Can the bot recognize that? Also, “…”
  • v
    Can I use that instead of a regular period?
  • i
    dont know about "..." but my rick model had stutters and the model works, only issue is that im pretty sure it cant generate the stutters i guess
  • v
    Hm, interesting. Yet again, thanks for the advice
  • s
    well then, the rec room stunt runner announcer isn't possible then
  • i
    yeah most likely wont
  • s
    that's why I wanted to get rid of the echo/reverb
  • s
    but with that, reverb I don't know if it can be removed without ruining the audio quality
  • s
    I've tried to remove some of the reverb. I don't know if would work still.
  • v
    I’d say that the reverb is subtle
  • v

    https://youtu.be/fhytVuLIFCQ▾

  • v
    Example
  • e
    what de-verbing solutions have you tried? this is still a huge challenge for DSP programmers, but there is iZotope RX and Zynaptiq Unveil
  • i
    Here's a suggestion, buy an external harddrive.
  • a
    bruh
  • a
    yes, using talknet atm
  • a
    yo this fucks up pitch 😂
  • u
    imagine gman sung
  • a
    ehhhh..... it works i guess...
  • u
    wait is that a website?
  • u
    localhost
  • u
    its called voice cloning app i think
  • u
1...279280281...1068Latest