Mistral Voxtral 3: directable text-to-speech in 35+ languages and cheaper transcription
Mistral released Voxtral 3, a family of speech models with a text-to-speech model you can direct for emotion, pacing and emphasis, and new transcription models with speaker labels. The TTS costs $13 per million characters and can adapt a custom voice from 30 minutes of audio.
What it does
Voxtral Mini TTS 3 speaks 35+ languages and 50+ dialects, follows plain-language instructions for emotion, pace and emphasis, and accepts phoneme hints for names. Transcribe 3 covers 50+ languages with diarization (who spoke when) in Mini, Small and Realtime versions.
Worth knowing
Transcription costs $0.0035 per minute (Mini), $0.0040 (Small) and $0.008 (Realtime). These are API models with self-deployment for enterprises, not open weights, so it is not tagged local. A Pro Voice option builds a custom voice from about 30 minutes of recordings.
Try it yourself
Read this line slowly and quietly, like a tired detective at 3 a.m., with a short pause before the last word: "[your line]"