Fish Audio S2 – Expressive AI voices controlled by natural language

Summary: Fish Audio S2 is an open-source text-to-speech model that enables users to direct voice expressions using natural language cues. It supports multi-speaker dialogue generation in one pass and produces realistic voices in over 80 languages.

What it does

Fish Audio S2 lets users embed emotion and expression tags like [whisper] or [laughing nervously] within text to control speech output. It generates multi-speaker dialogues and expressive voices with natural flow.

Who it's for

Developers and creators needing advanced, customizable TTS for multi-lingual, expressive voice applications.

Why it matters

It solves the challenge of producing natural, emotionally nuanced speech with directable voice cues in multiple languages.