"Apply Top Hit" trusts one source's first result, which is how a seven-minute remix ends up tagged with the three-minute single's album and year. Tag Verification asks several sources independently and only proposes a change where they agree — agreement is a far stronger signal than ranking.
It runs at three levels, all producing the same review screen: a free cross-source consensus that needs no account, Apple's on-device AI, and a cloud model you bring your own key for. You pick the trade-off; nothing is written until you review it.
What it does
- Cross-source consensus (free). Asks iTunes, MusicBrainz, Deezer, Discogs and Wikipedia independently and proposes a change only where they agree. No account, no key, works on every Mac.
- Wikipedia for the original album. Wikipedia is the most reliable source for the record a song first appeared on, where the catalog APIs return the single or a later hits compilation. It gets the final say on the album.
- Apple on-device AI (free, private). On macOS 26 with Apple Intelligence, the on-device model reads the same database results and judges which one matches the file. Nothing about your library leaves the Mac.
- Cloud AI — Claude, OpenAI, or your own local model. Claude can also search the web, for the bootlegs, edits and white labels the databases don't carry. Prefer OpenAI, OpenRouter, Groq, Mistral, DeepSeek, or a model you're already running locally under Ollama or LM Studio? Point it at any of those instead — billed to your own account, or free if it's running on your Mac.
- Fills a missing year from the file name. When the ID3 tag has no year but the file is named “… (2019)”, the year is offered instead of left blank — a database year still wins when one is found.
- Keeps the artist out of the title. “Justice - D.A.N.C.E.” becomes “D.A.N.C.E.”, while DJ version markers like (Extended Mix) and (Clean) are kept.
- Knows an edit from a mix. Candidates whose length cannot match the file are discarded before anything is counted, so a mix's album and year are never written onto a radio edit.
- You review before anything is written. Every proposal shows the current tags, the change, a confidence score and the source it came from. Applying writes to the files and Serato together, with a backup taken first.
How to use it
-
Select tracks and an engine
In Tracks & Tags, select what you want checked and choose consensus, on-device, or cloud.
-
Run it
A free offline check flags the tracks that actually look wrong first, so a paid or on-device run can be narrowed to just those.
-
Review the proposals
Each card shows the current tags, the proposed change, a confidence and the source. Pre-checked proposals are the confident ones.
-
Apply
Selected changes are written to the audio files and Serato together, read-back verified, with a snapshot taken first. Serato has to be closed.
The cross-source consensus is free and works everywhere. The on-device tier needs macOS 26 with Apple Intelligence; the cloud tier needs your own API key for Claude or an OpenAI-compatible provider — or a local model, which is free — and is billed to whichever account you use.
Related
Tracks & Tags
Browse the whole library, fix metadata in place, and fill thousands of empty fields in one pass.
Find Duplicates
Catch the same song saved five times, even when every copy is named differently.
Rename From Tags
Consistent filenames across your whole library — and Serato still finds every one of them.
Try it on your own library
Free and open source. Takes a backup before it touches anything.