AI voice company Fish Audio just closed a $52 million seed round co-led by Coreline Ventures and Today Capital. The startup started as the open-source project Fish Speech, offering text-to-speech, voice cloning, and real-time voice agents.
Fish Audio is only a year old and has grown from 3 to 22 employees. The platform has over 8 million users and reached $21 million in annual recurring revenue. The original model was trained by former NVIDIA researcher Shijia Liao on a single RTX 4090.
Its latest flagship model, S2.1 Pro, is not open-weight — it's only available via API. The company claims it can clone a voice from just 5 seconds of audio and supports over 83 languages. It's twice as fast as Cartesia and roughly one-sixth the price of ElevenLabs. The API is free through the end of August.
The funding will go toward developing audio understanding and voice-to-voice models. Companies like HeyGen, LiveKit, Retell, Sanas, and OpenArt are already using it in production. But the platform still faces complaints about unauthorized voice cloning of voice actors. Fish Audio says it has reduced the takedown process to about 3 minutes, but voices can still be used until a complaint is filed.