• cbarrick@lemmy.world
    cake
    link
    fedilink
    English
    arrow-up
    25
    arrow-down
    2
    ·
    1 month ago

    Fine tuning a general TTS model on a specific custom voice doesn’t require as much data as you think it does.

    The hard part is building the foundational model that can be easily fine tuned. And OpenAI has already done that.