Fish Audio
Expressive TTS and voice cloning in 80+ languages, with open weights
Overview
Fish Audio offers expressive, emotion-tagged text-to-speech and voice cloning from about ten seconds of audio in 80+ languages, via a studio and an API priced per UTF-8 byte, with open-weight S2 Pro under a research licence. Free personal plan; Plus from $15 a month, or $11 billed annually.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 8,000 credits a month (~7 min)
- 500 characters per generation
- 3 public voice slots
- 250,000 credits (~200 min)
- 10 private voice slots, Voice Design
- 1 professional voice slot
- Commercial use
- 2,000,000 credits (~1,620 min)
- 3 team seats
- 5 professional voice slots
- 7-day money-back guarantee
- 25,000,000 credits (~6,250 min)
- 10 team seats
- 15 professional voice slots
- Free s2.1-pro-free tier to 30 Nov 2026
- Speech-to-text $0.36 per hour
- Zero data retention
- On-premise deployment
- SOC 2 compliance
What you can produce with Fish Audio
- Voice cloning from about 10 seconds of reference audio
- Emotion-tagged TTS across 80+ languages incl. Malay, Indonesian, Thai and Vietnamese
- Cross-lingual cloning
- Developer REST API ($15 per million UTF-8 bytes)
- Open-weight S2 Pro under a research (non-commercial) licence
- Speech-to-text and Voice Design
- Inference in the US and Tokyo for enterprise
Fish Audio is one of the strongest challengers to ElevenLabs: expressive speech with inline emotion control, voice cloning from about ten seconds of audio and support for more than 80 languages, including most Southeast Asian ones. Its API is priced simply at $15 per million bytes of text, a free S2.1 Pro API tier runs until the end of November 2026, and consumer plans start at $11 a month billed annually. A $52 million raise and enterprise customers such as HeyGen show momentum. Two cautions: its "open-source" models carry non-commercial or research licences, and its own claim of first place on the TTS Arena is out of date. Comparisons that call it eleven times cheaper than ElevenLabs mix bytes and characters. A strong choice for developers and creators wanting multilingual voices at lower cost; check the licence before using open weights commercially.
What we found
Fish Audio makes expressive text-to-speech and voice cloning, sold through a consumer studio and a developer API by Hanabi AI, a Palo Alto company led by chief executive Rissa Cao with co-founder and chief scientist Shijia Liao. Its newest model, S2.1 Pro, released in June 2026, covers 83 languages — including Malay, Indonesian, Thai, Vietnamese, Tagalog, Burmese and Khmer — with fine-grained inline tags for emotion and delivery, and clones a voice from about ten seconds of clean audio. Fish says background music, reverb or overlapping speakers in a sample degrade the clone. In July 2026 it raised a $52 million seed round led by Coreline Ventures and Capital Today and reported $21 million in annual recurring revenue, eight million users and a team of 22; customers include HeyGen, Retell, Sanas and LiveKit.
Its models are open-weight rather than open source in the usual sense. S2 Pro, trained on more than ten million hours of audio, is published on Hugging Face under the Fish Audio Research License, which requires a separate licence for commercial use; older Fish-Speech and S1-mini weights are non-commercial. S2.1 Pro is available only through the paid API. Rankings have moved: Fish's documentation still claims first place on the TTS Arena, but the live leaderboard puts S2 at 13th and S1 at 22nd, while Artificial Analysis ranks S2 Pro second among open-weight models.
The free plan gives 8,000 credits a month, about seven minutes of speech, for personal, non-commercial use. Plus costs $15 a month, or $11 billed annually, for 250,000 credits, private voices, voice design and commercial use; Pro costs $100 ($75 annually) for two million credits and three seats; Max costs $999 ($749) for 25 million credits. The API charges $15 per million UTF-8 bytes of text, with a free S2.1 Pro tier under fair use until 30 November 2026. Enterprise adds zero data retention, on-premises deployment and SOC 2, with inference in the US and Tokyo. Claims that it is many times cheaper than ElevenLabs usually compare unlike units, so test with your own text.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Fish Audio's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Fish Audio unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Fish Audio directly →
Spotted something out of date? Suggest an update →
Related reading
Alternatives to Fish Audio
More in Video & Audio
