Fish Audio has open-sourced Drama 3, which is based on Qwen3.5-35B-A3B and features precise voice control.

Beating AI Express: AI audio company Fish Audio has officially open-sourced the weights of its Drama 3 model, allowing developers to download and deploy it independently from Hugging Face.

Drama 3 launched a preview version back in September, and now it fully opens its web platform, mobile app, and API, alongside releasing model weights, inference code, and technical reports. Built on the Qwen3.5-35B-A3B language model, Drama 3 has a total of 35 billion parameters, with only around 3 billion ac

@tonmelonfest