Fish Audio Launches Drama 3 for Granular AI Voice Direction

The platform aims to solve the rigidity often found in synthetic audio, where minor errors previously required full clip regeneration. Drama 3 introduces audio inpainting, allowing users to isolate specific segments for correction while maintaining the rhythm and mood of the surrounding audio. This capability extends to multi-character scenes, enabling developers to generate interactive dialogue where voices respond to one another within a single, continuous take.

"Voice AI has gotten very good at generating speech. The next step is directing a performance," said Rissa Cao, co-founder and CEO of Fish Audio. The model supports long-form consistency across English, Chinese, Japanese, and Korean, ensuring that character traits remain stable throughout extensive scripts. Shijia Liao, co-founder and chief scientist, noted that the model's architecture analyzes the context on both sides of a target edit to ensure seamless integration of new words.

This launch marks a significant milestone for the company, which recently topped Ahrefs’ list of the fastest-growing AI firms. With over 15 million users and $21 million in annual recurring revenue, Fish Audio is positioning Drama 3 as a professional-grade tool for industries ranging from animation and gaming to audiobooks. To encourage adoption, the company is offering an 80% discount on API usage for the first 30 days following the release.