NeptuneHub/AudioMuse-AI v0.3.4-beta

AudioMuse AI v0.3.4-beta: Analysis and Cluster on Embeding is now possible!

on GitHub

Release Date: June 19, 2025

Summary

AudioMuse-AI v0.3.4-beta introduces significant upgrades to analysis and playlist generation, with a major focus on experimental clustering using embeddings — now available in Advanced Configuration Mode.

⚠️ Note: To use clustering based on embeddings, you must re-scan your library to generate and store embeddings in the database. Standard (basic) clustering will continue to work with your existing database.

Experimental Feature: Clustering with Embeddings

You can now enable clustering directly on the embedding vectors via the advanced configuration settings. This unlocks deeper, more accurate grouping of tracks based on their AI-learned representation.

Suggested Configuration:

• Enable embedding-based clustering in Advanced Mode

• Set:

  • check Use Embeddings for Clustering:
  • Iterations: 10,000
  • Algorithm: GMM
  • PCA Components Min: 0
  • PCA Components Max: 199

This feature is still experimental. Results depend on proper configuration — tune parameters carefully for best performance.

🔧 Other Changes & Improvements

Instant Playlist Enhancements

• Improved prompt handling:

  • Special characters in artist/title (e.g., ') are now auto-escaped to prevent errors.

• New chat.html page:

  • Automatically retries the AI call up to three times when no or invalid results are returned.

Analysis Module Upgrades

• Now saves both:

  • Embedding vector (200-element array)
  • Feature vector

Clustering Updates

• AI-generated cluster names are now shorter to allow more creative naming.
• Experimental embedding clustering is currently only available in Advanced Mode (see above).

Miscellaneous

• Minor fixes and performance improvements.

What's Changed

• devel → main by [@NeptuneHub](https://github.com/NeptuneHub) in [#13](#13)

Full Changelog: [Compare v0.3.3-beta...v0.3.4-beta](v0.3.3-beta...v0.3.4-beta)