Release Date: July 20, 2026
AudioMuse AI v3.0.3 significantly improves Unique Catalog song deduplication, reducing false positives where different tracks were previously marked as duplicates.
Key improvements include:
- Stricter duration matching, reducing the allowed difference from 7 seconds to 1 second.
- Folder-aware validation, assuming songs in the same folder belong to the same album and are therefore different versions rather than duplicates.
- Chromaprint integration for more accurate audio fingerprinting.
Deployment behavior:
- New deployments immediately use Chromaprint to deliver more precise deduplication.
- Existing deployments begin collecting up to 1,000 album Chromaprints per analysis round (approximately 10–20 minutes per round). Once enough fingerprints are available, deduplication can be manually triggered from the Cleaning functionality.
Important
After the update we suggest to:
- wait that flask container start, it will take around 5-10 minute and unmap the song marked as duplicate for error: it will generate some orphan song. Are not real orphans, are song with missing data;
- run a cleaning task setting to true the flag to clean the catalogue: it will delete this orphan from the database bringing the database to a clean state;
- Run an analysis, it will re-analyze the false duplicate.
Depending on how much song was initially marked as duplicate for error this activity can take few hours or more.
For discussion about duplicate and orphan song please take a look here: #770 (comment)
The Cleaning functionality has also been enhanced with a new frontend option to clean the Shared Catalog by removing orphan songs that are no longer tracked by any server.