Fish Audio launches Drama 3 to direct AI voices and edit a single word

Fish Audio announced Drama 3 on October 13, a text-to-speech model that lets creators direct delivery in plain language, generate multi-speaker scenes, and repair a selected segment without regenerating the surrounding performance. It ships as the drama-3-preview model with 80% off API usage for 30 days; the company reports 15 million users and $21 million ARR.

• Single-word edits preserve voice, mood and rhythm on both sides
• Multi-speaker scene generation with language changes within a scene
• Phoneme-level pronunciation control for English, Chinese, Japanese and Korean
• API model drama-3-preview: 80% discount for 30 days, three free months