Release Date: June 11, 2025
🚀 Summary
In AudioMuse-AI v0.3.1-beta, we empowered the Clustering function with a new stratified approach. This approach helps the clustering function to create playlists also for genres for which you have a small number of songs.
The new clustering algorithm will sample a certain number of songs (editable with the percentile and min number of songs parameters) for each genre analyzed in your database. Then, a small part of them will be shuffled at each iteration. In this way, you will have a higher probability of creating playlists for the less represented genres.
In addition, the scoring function was improved, and the form to change the weights was added to the front-end. At the moment, the most tested combination is:
- Diversity score with a weight of 2
- Integrity score with a weight of 1
- All other weights set to 0 to disable them.
Further tests and improvements will be made in the next release.
This release will be done with the github option "Create a discussion for this release" with category "Q&A". Please use it for any question or suggestion on this or any other release. Your feedback is important.
What's Changed
- Devel -> Main by @NeptuneHub in #8
Full Changelog: v0.3.0-beta...v0.3.1-beta