Voxtral Mini TTS 3

Changes from previous version

Adds speech output to the Voxtral 3 family after Voxtral Mini Transcribe Realtime 3. Batch and realtime speech share one API price of $13 per million characters.

Release Summary

Voxtral 3 text-to-speech model, in public preview as voxtral-mini-tts-3-0. The same model card lists the realtime ids voxtral-mini-tts-realtime-3-0, and the announcement prices batch and realtime together at $13 per million characters. The post describes 35+ languages and 50+ dialects, with voice presets in 18 languages. The model card lists 9 languages, about 90 ms to first audio, 34 emotion tags, 2 pace tags, and five output formats including 24 kHz WAV. Pro Voice adaptation uses 30 minutes of audio. Text and audio in, audio out. The card omits a token context, so this entry keeps the 32k figure from the Voxtral Small Transcribe 3 card and records output as 0 because the API returns audio. Listening stays on Voxtral Mini Transcribe Realtime 3.

Timeline

Other Voxtral versions