Fish Audio Raises $50 Million in Funding to Build AI Voice Models

Palo Alto-based startup Fish Audio has raised $50 million in early-stage funding to build AI-driven voice models. According to ixbt.com, the funding round was led by Coreline Ventures and Capital Today, with contributions from a number of other venture funds. This was reported by Techcrunch.com reports .

The market for AI-generated voice models is currently growing rapidly. While creative professionals demand voices that are as expressive as possible, enterprises seeking to automate customer support and sales processes look for easy manageability. Fish Audio aims to meet all these requirements with its library of over 15,000 natural language control mechanisms.

The startup's short-term success

Since launching its operations last year, the number of users utilizing the startup's open-source and hosted versions has exceeded 8 million. Currently, the company boasts an annual recurring revenue of $21 million. The newly raised $50 million investment is planned to be used to further improve these metrics and expand infrastructure.

The Fish Audio project was originally founded as a small initiative by former NVIDIA researcher Shijia Liao. Dissatisfied with the lack of expressiveness in synthetic voices on the market, the specialist developed a new voice generation model using a single GPU and released it in an open-source format. The Fish Speech repository on GitHub currently has over 31,000 stars, actively used by independent developers, video game designers, and content creators.

Model lineup and enterprise solutions

Over the past year, the company has released five new models, including four speech generation models and one speech-to-text model. While the code for three of them has been published as open source, the latest S2.1 Pro model is available exclusively via a paid API. The company offers monthly paid subscription plans for creators and teams that include minute limits and voice cloning features.

There is also an enterprise API platform designed for organizations, with prominent companies such as HeyGen, Sanas, and Plaud utilizing its services. HeyGen uses the technology to give realistic voices to AI avatars, while gaming studios need expressive voices for their characters.

Copyright and automated solutions

To expand its voice library, the startup asked users to submit their voice samples and paid compensation if they were used. However, a few months ago, some creators complained that their voices had been uploaded without consent. To resolve this issue, Fish Audio's management fully automated the content removal process.

Now, authors can ensure that their voice is removed from the platform in less than 3 minutes by providing a short voice sample or a contract. Nevertheless, the issue remains relevant as the possibility persists for someone to upload another artist's voice without notification, allowing it to be used on the platform until the rightful owner files a complaint.