Fish Audio Launches Drama 3 Voice Creation Model

Fish Audio, providers of artificial intelligence platforms building expressive text-to-speech, speech-to-text, and voice cloning, has launched Drama 3, a text-to-speech model that ets developers, studios, and creators direct voice performances instead of simply generating them. Users describe the desired tone, pacing, and character in plain language, then shape the performance with line-by-line precision until it's right.

Drama 3 is for a wide range of applications, including audio dramas and interactive media, audiobooks, multilingual dubbing, games, animation, AI companion experiences, and content creation platforms. A drama studio can direct a two-character argument as a single scene, an audiobook narrator can shift from calm to trembling mid-sentence, and a game team can fix even a single word of dialogue without re-rendering the clip. S

"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, in a statement. "Drama 3 is our most emotive and controllable model yet. It puts creators in the director's chair: Describe what you want in plain language, listen back, and adjust only the moments that aren't working. That matters most in long-form projects like audiobooks and multilingual dubbing, where one wrong word shouldn't mean regenerating a whole chapter."

Drama 3 gives teams four new ways to shape a performance:

• Natural-language direction: Describe a performance in plain language and change a sentence mid-line so it can start calm and turn panicked without a new take.
• Multi-speaker dialogue: Generate a whole scene in one pass while each character keeps its own voice and responds to what the other just said.
• Audio inpainting: Select the part of a performance to edit, update the instructions, and Drama 3 regenerates only that segment, keeping everything around it intact.
• Long-form and multilingual consistency: Keep the same voice and character across long scripts, switch languages mid-scene, and control pronunciation at the phoneme level in English, Chinese, Japanese, Korean, and more. Unlike models that stitch separate requests together for long-form content, Drama 3 holds one performance across the full script.

"Drama 3 adopts a new architecture that allows for better control and instruction following," said Shijia Liao, co-founder and chief scientist of Fish Audio, in a statement. "When you change a single word, Drama 3 listens to the audio on both sides of it and makes the new word fit. Same voice, same mood, same rhythm. It sounds like a small thing, but it was one of the hardest parts of the model to get right."

Drama 3 caps a run of launches for Fish Audio, which marked its first anniversary this summer. In the past few months alone, the team has shipped a speech-to-text model with built-in speaker diarization; Fish Creative, a single workspace for professional content creators; and mobile apps for iOS and Android. Drama 3's predecessor, S2.1 Pro, became the most-requested speech model on OpenRouter within three weeks of launch.