Fish Audio Drama 3 brings more emotion and control to AI voice generation
AI-generated voices are getting a major performance upgrade. Fish Audio has introduced Drama 3, a new text-to-speech model built to make synthetic dialogue sound more natural, expressive, and emotionally convincing.
Launched on October 13, Drama 3 is designed for creators who want more than a flat voice reading lines from a script. The model allows users to guide vocal delivery with simple, everyday language. That means creators can tell an AI voice to sound nervous, angry, excited, hesitant, exhausted, or calm without needing complicated audio tools or multiple recording sessions.
One of the standout features is the ability to change emotion in the middle of a sentence. A character can start a line softly, grow frustrated, and end in a whisper, all within the same piece of dialogue. This gives writers, game developers, video creators, audiobook producers, and audio drama teams more control over how a scene feels.
Instead of choosing from basic voice presets, users can place direction directly into the script. For example, a line can include instructions for pacing, tone, emphasis, or mood. Drama 3 can apply those changes to a single word, a short phrase, or a longer passage, making it easier to fine-tune a performance without altering the entire recording.
The model also supports multi-character conversations, which could make it useful for podcasts, animated projects, game dialogue, interactive stories, and other voice-driven content. With support for 83 languages, Drama 3 is also aimed at creators working across international markets.
Fish Audio says Drama 3 scored 93.2% on the Artificial Analysis Pronunciation Robustness benchmark. That is higher than the company’s previous S2.1 Pro model, which scored 88.1%. While strong pronunciation does not automatically mean every performance will sound fully human, it does suggest the new model is better at handling difficult words, names, and varied speech patterns.
The company is also working on two editing tools that could make AI voice production more flexible. Interactive Editing will let users improve a performance through follow-up instructions, similar to giving direction to a voice actor. Audio Inpainting will allow creators to replace individual words without regenerating an entire audio clip. Both tools are currently in limited invitation-only testing.
Drama 3 is available through the Fish Audio website, mobile apps, and API. The company is also offering downloadable model weights, though commercial use comes with licensing restrictions.
For early users, Fish Audio is offering discounted API pricing until November 12. During the promotion, access costs $10 per million characters, down from the regular $50 rate. New creator subscriptions are also available at 50% off for a limited time.
With Drama 3, AI voice generation is moving closer to performance-based storytelling. For creators who need expressive narration, believable characters, and more control over vocal emotion, this model could make synthetic voices feel less robotic and much more dramatic.
