Mini toolsFish Audio S2
By Fish Audio
Model family: Fish
Fish Audio S2 is a dual-autoregressive TTS system trained on over 10 million hours of audio across roughly 50 languages, aligned with reinforcement learning, and built to support word-level, inline natural-language control tags (for example [laugh], [whispers], [super happy]) for prosody and emotion. It also supports native multi-speaker, multi-turn generation and is released together with a production streaming inference stack (SGLang-based), model weights, and fine-tuning code.
Audio
Overview
Fish Audio S2 is Fish Audio’s latest text-to-speech model designed for natural, emotionally rich speech generation with fine-grained prosody control and native multi-speaker dialogue.
Last updated: July 9, 2026
