Fish Audio
Fish Audio S2
Fish Audio S2-Pro is Fish Audio's next-generation expressive text-to-speech model for natural voice generation, open-ended emotion tags, multi-speaker dialogue, voice cloning, and 80+ language workflows.
Fish Audio currently recommends s2-pro for new projects. S2-Pro adds natural-language bracket control such as [whispers sweetly] or [laughing nervously], supports multi-speaker dialogue, covers 80+ languages, targets 100ms time-to-first-audio, and ships with an open-source SGLang-based serving stack. The previous s1 model remains available for existing integrations and is still useful when a workflow depends on its parenthesis-based emotion syntax. On an Astorie SEO page, Fish Audio is positioned as an expressive voice foundation model to compare against ElevenLabs, Minimax Speech, and other TTS systems; it is not wired into Astorie's production generation menu unless a runtime integration is added separately.
Fish Audio S2 Variants
| Variant | Description |
|---|---|
| Fish Audio S2-Pro | Recommended current model with bracket-style natural language control, multi-speaker dialogue, 80+ languages, and open-source serving. |
| Fish Audio S1 | Previous 4B-parameter model with parenthesis-based emotional control, kept for existing integrations. |
Capabilities
Best For
- Expressive voiceovers with natural-language emotion and delivery cues
- Multi-speaker dialogue and character audio prototypes
- Developers who want open-source serving and fine-tuning options
- Voice cloning workflows that need direct reference voices
- Multilingual audio experiments across broad language coverage
Strengths
- S2-Pro is the recommended current Fish Audio model for new projects
- Open-ended bracket control is more flexible than a fixed list of emotion tokens
- Multi-speaker dialogue is exclusive to S2-Pro in the Fish Audio TTS API
- 80+ language support with automatic language detection
- Open-source model and serving stack reduce vendor lock-in for teams that can self-host
Limitations
- Fish Audio is SEO-only in Astorie until a production runtime integration is added
- S2-Pro uses a different bracket syntax than S1, so old parenthesis prompts may need rewriting
- Self-hosting and fine-tuning require infrastructure and audio engineering effort
- Voice quality still depends on clear reference voices, transcripts, and prompt discipline
Tips & Best Practices
Use Fish Audio S2 on Astorie
Connect Fish Audio S2 with video, image, script, and music nodes on Astorie's infinite canvas. No GPU required — start free.
Get Started FreeFrequently Asked Questions
What is the latest Fish Audio TTS model?
Fish Audio currently recommends s2-pro for new projects. It adds natural-language bracket control, multi-speaker dialogue, 80+ languages, low time-to-first-audio, and an open-source serving stack. S1 remains available for existing integrations.
Is Fish Audio S2 available as an Astorie generation model?
This page is an SEO and comparison page. Fish Audio is not added to Astorie's production audio generation menu by this change; that would require a separate runtime provider integration, billing, UI controls, and webhook handling.
How is Fish Audio different from ElevenLabs?
Fish Audio S2 emphasizes open-source serving, flexible bracket-style control, and self-hosting options. ElevenLabs emphasizes a mature hosted voice ecosystem, Eleven v3 expressiveness, Multilingual v2 stability, and text-to-sound effects.