Stability AI releases open-weight Stable Audio 3.0 models

Stable Audio 3.0 was earlier than today's sweep, but it was a clear missed model-release story. Stability AI says the model family is trained on licensed data and designed for artistic experimentation with open weights. Small and Medium are available on Hugging Face, while Large is available through the Stability AI API and self-hosting for enterprise deployments. The system supports variable-length generation, full song composition on portable devices, audio-to-audio transformation through natural-language prompts, 44.1kHz stereo output, and tracks up to roughly six minutes. The release matters because open-weight audio models are rarer than open image or text models, especially when training data licensing is a central claim. Watch license limits, artist adoption, quality compared with Suno and ElevenLabs, and whether local audio generation becomes practical for indie creators.

Key details: Stability AI, Stable Audio 3.0, May 20, 2026, open-weight music models, licensed training data, Stable Audio 3.0 Small, Stable Audio 3.0 Medium, Hugging Face.

Continue swiping for more AI Brief stories.