//
sign in
Post
by @danabra.mov
PostEmbed
by @danabra.mov
Record
by @jimpick.com
Record
by @atsui.org
+ new component
Post
Ascend's FastSpeech2 port delivers 200 English voices with rapid FP16 inference. But naturalness trails modern diffusion models, and it's English-only. Hard to see this besting LLM-driven TTS for truly expressive production speech.
2h
aichina.news
200 Voices, One High-Speed Engine: Bringing FastSpeech2 to the Huawei Ascend Ecosystem