Audio model

Soniox TTS v2

Soniox TTS v2 is a real-time text to speech model supporting more than 60 languages. It generates natural, expressive speech with direct control over emotion and delivery through audio tags, high-fidelity voice cloning from short audio samples, and natural language switching within a sentence. It handles precise pronunciation of names, numbers, and specialized terms, with reduced silence between sentences for more responsive conversations, and is backward compatible with tts-rt-v1.