News Summary

French AI company Mistral has unveiled Voxtral, a new open-source speech model aimed at elevating the capabilities of voice AI applications. This model supports nine languages and is designed for real-time performance with applications in voice AI assistants and enterprise use cases such as customer support. Mistral’s model, Voxtral TTS, can adapt custom voices with minimal samples and maintain characteristics like accents and intonations even when switching languages. This release is part of Mistral’s broader strategy to offer a full suite of voice products for enterprises, with the open-source nature allowing for customization and adaptation by users. Additionally, Mistral released Voxtral Mini, optimized for efficient transcription and suitable for sensitive audio processing. These developments position Mistral as a significant competitor to major industry players like ElevenLabs, Deepgram, and OpenAI, leveraging open-source advantages to expand its reach.