The AI voice generation sector has once again seen a major funding round. Palo Alto, USA-based Fish Audio announced it has completed a $50 million seed funding round, led by Coreline Ventures and Today Capital. The company stated that one year after its platform's launch, the total number of users for both open-source and managed models has exceeded 8 million, with annual recurring revenue reaching $21 million.
Users and revenue have reached a certain scale.
Fish Audio primarily provides speech generation and speech-to-text models for creator teams and enterprise clients. The company claims its products support over 15,000 natural language controls for adjusting speech style, expressiveness, and controllability, covering scenarios such as content creation, gaming, customer service, and sales.
Over the past year, Fish Audio has released five models, four of which are speech generation models and one is a speech-to-text model. The company has open-sourced three of the speech generation models, while the latest S2.1 Pro is only available through a paid API.
- The total number of users exceeds 8 million.
- Annual recurring revenue is $21 million.
- The Fish Speech project on GitHub has received over 31,000 stars.
Serving both creators and enterprise clients
Fish Audio offers a monthly subscription plan that provides individual creators and teams with a fixed amount of voice generation credits, and also offers a voice cloning feature. The company also sells enterprise-level APIs and platform services, stating that organizations such as HeyGen, Sanas, and Plaud are already using its products.
Cao, co-founder and CEO of the company, said that different companies have different requirements for voice models. AI digital human companies value realism, game studios value character expressiveness, while voice agent companies focus on low latency and naturalness in calls.
Voice authorization remains a key issue in the industry.
Fish Audio initially expanded its voice library by accepting user-uploaded voice samples and paid submitters when their voices were used. However, several months ago, a creator claimed that their voices were uploaded to the platform without their consent.
The company stated that it previously had a DMCA-based removal process, but the processing speed was slow. This process has now been automated. Creators only need to submit a short audio sample or a relevant contract to prove ownership of the sound, and the platform can complete the removal within 3 minutes.
However, this mechanism is still mainly used for post-event processing and cannot completely prevent others from uploading other people's voices without authorization. Investor Coreline Ventures also stated that for a community-driven model to be sustainable in the long term, creators must have sufficient trust in the platform, and authorization, transparency, attribution, and clear licensing terms need to be incorporated into the product design.
The funds will be invested in the research and development of new models.
Fish Audio stated that in its early stages, focusing on open-source projects and creator subscriptions, the company operated efficiently and did not have a strong need for external funding. However, as investor interest grew and the company aimed to develop more advanced models and expand into the enterprise market, fundraising became the next step.
The company plans to launch an audio understanding model this year and is developing a speech-to-speech model. The AI speech generation market is currently highly competitive, with companies like ElevenLabs, WellSaid, Cartesia, Speechify, Async, and Krisp vying for creator and enterprise budgets. Investor 359 Capital believes that finer-grained developer control and more efficient training costs will be key factors for Fish Audio in competing with large AI labs.
