From the source
Together AI is now serving Qwen3 TTS, Qwen's open text-to-speech model, on its Serverless Inference and dedicated endpoints at $20 per 1M characters.
- The model separates who is speaking from how they speak, so a chosen voice keeps its identity while a plain-language instruction such as "excited and upbeat" changes its delivery.
- Nine preset voices are offered, and every one of them speaks Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish and Italian.
- Together AI says its inference team tuned the serving stack so speech starts streaming fast enough for real-time conversation, with first audio in as little as 97ms.
- Developers can stream over HTTP with server-sent events or use a WebSocket API when the text arrives in pieces, and the output is 16-bit mono PCM at 24 kHz.
- Qwen recommends using a voice in its native language for the best quality, and word timestamps are not supported for this model.






