Fish Audio Secures $52 Million Seed Round and Launches S2.1 Pro, a Model That Clones Voices from 5 Seconds of Audio and Adjusts Emotion with Tags
Fish Audio has completed a $52 million seed round financing and launched the advanced voice AI model S2.1 Pro.
This model can clone voices from 5 seconds of audio and adjust emotions with tags, with a latency of less than 300 milliseconds, costing one-sixth of ElevenLabs. The 22-person team has achieved $21 million in annual recurring revenue.
Event-driven voice AI startups are accelerating, with funding flowing towards low-cost, high-performance models, putting pressure on traditional voice synthesis manufacturers while benefiting the developer ecosystem.
Source: Public Information
ABAB AI Insight
Fish Audio originated from a GitHub open-source project by a former NVIDIA researcher, expanding from single-card training to multiple models and commercial APIs within a year, serving clients like HeyGen and Retell while maintaining open-source and free developer access.
The capital path shows that the seed round was led by Coreline and others, with resources invested in model iteration and enterprise sales, motivated by capturing the creator and customer service markets with emotional control and extremely low latency, specifically through community libraries and low-price strategies to quickly acquire customers.
Similar to ElevenLabs' path from open-source to high valuation, the current voice generation industry is in an expansion phase from basic synthesis to controllable emotion and real-time streaming.
Essentially a technological replacement, the mechanism significantly reduces customization costs through short sample cloning and natural language tags, enabling small and medium teams to deploy high-quality voice agents.
ABAB News · Cognitive Laws
Open-source projects can be quickly commercialized
Five-second cloning rewrites the voice barrier
Low price and high quality are customer acquisition levers