💛 Donate to help keep AudioMuse-AI growing. Your support covers AI subscription for development (~€220/month), homelab electricity for testing (~€20/month), and makes the project sustainable.
Release Date: August 13th, 2026
AudioMuse AI v3.3.0 introduces the new Hyperbolic Explorer, allowing you to discover songs and playlists that are further away from your usual music. The integrated tree view also lets you create playlists directly from your discoveries, including playlists ordered by genre, while still being based entirely on sonic similarity.
This release also adds an experimental -nvidia-arm image with NVIDIA support on ARM, especially designed for DGX Spark and other GB10-based machines.
Finally, you can now mark songs as not analyzable during the analysis process. These songs will be skipped in future analysis runs, saving analysis time. Their status can also be checked directly from the dashboard.
Other minor fix was added especially on the way of which the idle ram is used on both flask and worker container.
Important
Hyperbolic Explorer to be used need the run of a new analysis. Running it with Number of Recent Albums equal to 0 is enough. It will create the ne hyperbolic index and the Hierarchy Directory tree.
Note
As usual, when updating, make sure to update the Flask container and all Worker containers at the same time. The application can run with one or more Worker containers, but they must all use the same version. Mixing different versions is not supported.
What's Changed
- Analysis dashboard log fix by @NeptuneHub in #847
- Analysis: skip song in error from previous run by @NeptuneHub in #850
- Add ARM64 (DGX Spark) NVIDIA image build workflow by @NeptuneHub in #854
- Hyperbolic Explorer + Process Fork fix by @NeptuneHub in #853
- cudnn_conv_algo_search TO HEURISTIC - #798 #783 by @NeptuneHub in #855
Full Changelog: v3.2.0...v3.3.0