Fish Audio Releases Drama 3: Turning TTS into Editable Audio, Re-record Just One Word If One Word Is Wrong

Beating AI News Flash: AI voice company Fish Audio has released a preview of its voice model Drama 3, focusing on more granular voice control. Users can directly describe tone, pacing, and character in natural language, without needing to piece together a bunch of emotion or audio tags.

Drama 3 can change voice and speaking style mid-sentence, and can also generate dialogue for multiple characters at once. If any part is unsatisfactory, users can redo just one word without regenerating the entire passage.

Fish Audio's previous generation S2 already supported natural language voice control. Drama 3 continues to refine that control, putting mid-sentence voice switching, multi-character generation, and localized redo into a single model. For game dialogue, audio content, and AI characters, post-production edits will be much more convenient.

Drama 3 is currently still in the Preview stage, and developers can apply for API testing access.