Fish Audio Drama 3 directs every word
Fish Audio launched Drama 3, an open-weight TTS model that lets developers direct tone, pacing, emotion, sounds, and word-level delivery using natural language. It supports multi-character scenes, 83 languages, and API access through the existing TTS endpoint. [Fish Audio launch post](https://fish.audio/zh-CN/blog/fish-audio-drama-3/)
Drama 3’s meaningful advance is controllable performance, not merely more natural speech.
- –Natural-language directions and scoped tags enable precise mid-sentence emotional changes.
- –Word-level audio inpainting can fix a line without regenerating an entire passage.
- –The existing API endpoint and `model: drama-3` header make adoption relatively low-friction.
- –“Open-weight” does not mean fully open-source: commercial use above the license threshold and hosted inference require Fish Audio permission.
- –Preview outputs can vary between takes, so production teams still need consistency testing. [Drama 3 product page](https://beta.fish.audio/drama-3/)
DISCOVERED
1h ago
2026-10-13
PUBLISHED
1h ago
2026-10-13
RELEVANCE
AUTHOR
FishAudio