Fish Audio S2 Pro

Fish Audio S2 Pro is an open-source text-to-speech model supporting 80+ languages with natural-language prosody and emotion control via bracketed style cues. It features a dual-path architecture (4B Slow AR + 400M Fast AR) enabling multi-speaker dialogue, voice cloning, and streaming inference with ~100 ms time-to-first-audio latency. Source: supabase.