KyutaiKyutai TTS
Kyutai TTS 0.75B English Public
Kyutai TTS 0.75B English Public converts a stream of text into speech. It belongs to Kyutai's Delayed Streams Modeling family and can begin producing sound before all text has arrived.
Research release
The release has its own audio configuration
This English model was trained on public data for reproducible research. It uses 16 audio tokens per frame. The release does not support reducing that token count at inference.tts-0.75b-en-public: Kyutai model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | kyutai/tts-0.75b-en-public |
| Parameters | 300M backbone and 450M depth transformer.tts-0.75b-en-public: Kyutai model card |
| Weight license | CC BY 4.0.tts-0.75b-en-public: Kyutai model card |
| Text/audio offset | 16 frames, equivalent to 1.28 seconds at 12.5 frames per second.tts-0.75b-en-public: Kyutai model card |
Related models
A streaming neural audio codec that compresses speech into tokens and reconstructs sound from them.
Launch date not confirmedResearch releaseDownloadable