French AI company Mistral AI has launched Voxtral TTS, a new open-source text-to-speech model designed for voice assistants and enterprise applications like customer support and sales. The model supports nine languages including English, French, German, and Arabic, and can create a custom voice from less than five seconds of audio. It maintains voice characteristics like accent and intonation when switching between languages, making it suitable for dubbing and real-time translation.
With latency as low as 90 milliseconds, Voxtral TTS is reportedly faster and costs a fraction of other market models.