Release Date: July 20, 2026
AudioMuse AI v3.0.4 continues to improve the migration experience by fixing Out of Memory (OOM) issues that could occur with some large libraries and by automatically cleaning up false orphan songs created during migration (now no orphan will appear anymore due to a migration).
This release also brings improvements to AI Naming for clustering and fixes a bug in Song Alchemy Anchor where the excluded song list was not correctly saved as part of the anchor configuration.
Important
From v3.0.3 At the end of the analysis (at 99%), you may notice a final Chromaprint stage for songs imported from legacy versions. During this stage, AudioMuse AI calculates Chromaprint fingerprints for songs that have already been analyzed.
By default, each analysis processes 1,000 previously analyzed albums. This gradually backfills the entire library. You can change this behavior with the CHROMAPRINT_BACKFILL_ALBUMS_PER_RUN Setup Wizard parameter:
- Increase the value to complete the backfill faster.
- Decrease the value to reduce the impact on each analysis run.
Once every song has a Chromaprint fingerprint and a Cleaning functionality task is executed (with Also delete orphans from the catalogue enabled), Chromaprint will provide an additional safeguard against duplicate detection. It may also be used as the foundation for a centralized song ID and future implementation. New track already have the chromaprint analyzed inline.
As usual we suggest to do a backup before the update. If you use the latest image that auto update, we suggest to do frequent backup