Release Date: November 30, 2025
AudioMuse AI v0.7.12-beta introduce important improvement in Instant Playlist based on AI and Clustering functionality.
The Instant Playlist now uses a new tool-driven AI system that focuses on execution plans instead of relying solely on the AI’s own music knowledge.
Previously, playlist quality depended heavily on the AI model’s internal understanding of music, sometimes leading to empty playlists or suggestions the user didn’t own. Now the AI simply decides how to use a set of powerful tools that query your actual library and APIs directly. This makes results far more accurate and consistent, even on lighter AI models.
Clustering now includes GPU-accelerated implementations of K-Means, DBSCAN, and PCA. This gives real purpose to the -nvidia container image: previously, only the Analysis module benefited from GPU speed-ups—now the core clustering features do as well. Early tests show K-Means running up to 3× faster when executed on GPU.
IMPORTANT: to use this GPU clustering you need to set this env var USE_GPU_CLUSTERING at true, by default is false
Finally the Docker Model Runner–Ready docker compose configuration was added in the example of deployment, for whom want to deploy the AI model directly with Docker.
What's Changed
- [FIX BUG] 'tuple' object has no attribute 'replace' by @moutasem1989 in #200
- Add comprehensive unit tests for core modules by @tysoncung in #201
- Devel -> Main by @NeptuneHub in #204
- Give pointers as to how the music map works by @Sakrecoer in #207
- GPU Accelaration for Clustering by @rendyhd in #195
- Add Docker Model Runner–Ready Compose Configuration by @moutasem1989 in #194
New Contributors
- @Sakrecoer made their first contribution in #207
Full Changelog: v0.7.11-beta...v0.7.12-beta