Paris AI company Mistral releases Voxtral 3, a set of speech models and a voice agent SDK whose Pro Voice adaptation, Mistral says, reproduces "a speaker's distinctive character, emotional range, and speaking style across languages, using just 30 minutes of audio". The text-to-speech model covers more than 35 languages and 50 dialects and costs $13 per million characters, the transcription models handle more than 50 languages including all 24 official EU languages from $0.0035 a minute, and everything runs on Mistral's EU-hosted API or on a customer's own servers. The launch names contact centres and virtual assistants as uses, not dubbing or media.