Fish Audio released Drama 3, a text‑to‑speech model with a total of 35 billion parameters but only about 3 billion active at any time thanks to a mixture‑of‑experts (MoE) architecture. The model can follow plain‑language prompts and can generate speech in 83 languages. Fish Audio is making the model’s weights publicly available under a community license that caps revenue for commercial use.
Why it matters
Developers can now build multilingual voice applications without paying for a proprietary API, as long as they stay within the license’s revenue cap.
