Stability AI Updates: May 10, 2026

1. stable-audio-metrics receives updates

Stability AI. The stable-audio-metrics repository, which provides evaluation metrics for music and audio generative models, received commits on May 9. The toolkit is used internally and by the community to benchmark generative audio outputs against reference distributions. Source

2. stable-audio-tools repository updated

Stability AI. The stable-audio-tools package, which hosts generative models for conditional audio generation, received updates on May 8. The repository is the primary open source surface for Stability’s audio research and underpins downstream Stable Audio integrations. Source

3. ReLi3D relightable reconstruction code refreshed

Stability AI. The ReLi3D repository, containing code for relightable multi-view 3D reconstruction, was modified on May 8. The project targets photorealistic 3D asset generation with controllable lighting, an area Stability has been investing in for game and film pipelines. Source

4. marble repository updated

Stability AI. Stability’s marble Python repository received commits on May 8. The project is part of the company’s ongoing research code releases on GitHub. Source

5. diffusers fork synced

Stability AI. Stability’s fork of the Hugging Face diffusers library was updated on May 8. The fork tracks state-of-the-art diffusion model implementations and is used to stage Stability-specific patches and model integrations. Source