generated at
VALL-E
>TTSの非常にインパクトの大きい論文が出たので紹介.
>VALL-E: Zero-shot音声合成システム. MetaのEnCodecを取り入れ, LLMタスクとみなすことでTransformerの強みを活かしている。論文タイトルからも察することができるように❌DALL-E論文と同様の立ち位置であることを示す
>


とても自然だと思うwogikaze
>Looks like a community reproduction of VALL-E may come before the official release (no ETA or commitment from MSFT yet).
>
>We may be able to clone anyone’s voice to synthesize any speech on @huggingface soon 😮
>
>Link: https://t.co/sCYZ0PEOAL. Not lucidrains this time😄 Jim Fan
>
ほー非公式の実装がもう出たのかwogikaze