NeptuneHub/AudioMuse-AI v0.6.0-beta

AudioMuse AI v0.6.0-beta: Librosa and ARM support

on GitHub

Release Date: July 10, 2025

AudioMuse-AI v0.6.0-beta is here and it's a big step forward in compatibility and clustering accuracy. We've officially introduced support for ARM architecture and added a brand-new Spectral clustering method to give you even better organization and discovery.

Note: For this new version we recommend deleting all database tables and rescanning your library

Highlights

• Now Running on ARM!
  AudioMuse-AI now supports ARM-based systems thanks to Librosa integration (used in place of Essentia), making it usable on devices like Raspberry Pi, ARM servers, and other low-power systems.

• Librosa Is the New Default
  Song analysis now uses Librosa to read audio data, increasing compatibility across systems.
  TensorFlow continues to be used with the same model (.pb), ensuring no change in audio feature extraction.
  We recommend this version for most users moving forward.

• Essentia Version Still Available (if needed)
  A special -essentia version (:latest-essentia, :0.6.0-beta-essentia) is maintained for backward compatibility. Use only if you need the older setup for specific reasons.
  Switching between Librosa and Essentia require a full library rescan.

• New Clustering Algorithm: Spectral Clustering
  Discover songs in a whole new way with Spectral Clustering — offering a more nuanced organization of your music library.

• Clustering Fixes and Improvements
  Several fixes were applied to improve the stability and accuracy of the clustering process.

• Better Error Handling in Analysis
  Improved robustness with targeted fixes to reduce analysis errors and improve overall reliability.

• Enhanced Logging
  Logging has been upgraded across the board for easier debugging and deeper insight — especially in clustering, where you now get the full list of parameters logged clearly.

What's Changed

• devel-librosa -> main by @NeptuneHub in #22

Full Changelog: v0.5.0-beta...v0.6.0-beta