Fish Audio Secures $52M to Scale Expressive Voice AI

Co-founder and Chief Scientist Shijia Liao, a former NVIDIA researcher, launched the company after growing frustrated with the robotic quality of existing synthetic voices. By training models on a single gaming GPU, he created the open-source project Fish Speech, which garnered over 31,000 stars on GitHub. That initial momentum fueled the platform's current capability to clone voices from five-second clips in roughly 15 seconds, supporting over 83 languages with nuanced emotional control.

CEO Rissa Cao noted that the company’s growth stems from a commitment to human-sounding output that remains accessible to everyone from independent creators to large-scale enterprises. The startup's latest model, S2.1 Pro, outperformed leading competitors in 67% of blind listening tests, helping the firm secure partnerships with companies like HeyGen and Retell.

Coreline Ventures and Capital Today led the funding round, which the company plans to deploy toward expanding its audio-native stack, including voice-native large language models and speech-to-speech technology. To drive further adoption, Fish Audio is offering its S2.1 Pro model for free via API through the end of August as it scales its enterprise sales and developer tooling operations.