First, a point about naming- Auto Correct (on the main Action menu) is a workflow scripter. It should be renamed, IMO. Autocorrect tags from MB (on the Remote Actions submenu) attempts to add/correct tag metadata. It?s the latter I?m discussing here.
On the accuracy of Autocorrect tags, it is very good at identifying artist and song title. The few errors that occur here are probably due to errors in the MB or MusicIP databases that Jaikoz relies on. What?s more problematic is identifying the proper release for a song. A song may exist in promo, live or studio versions, with radio edits or remixes as additional variations. The song may be found on single, EP or album releases, best-of or various artist compilations, bonus track or special editions as well as a myriad of country specific offerings. Discriminating between these is a very difficult task. Acoustically, the song on the best-of album is often identical to the original studio album.
To properly tag the song, Jaikoz must rely on existing metadata (release title, track number, year, song length) or context (number of tracks in the album folder). In my experience, Jaikoz gets the release title correct about 90% of the time. It might be higher, it might be lower, but in large collections hundreds of songs will be misassigned. I?ve seen this happen for songs that were already properly tagged and grouped. If file name/hierarchy is based on tags, any change like this that slips through is difficult to correct.
http://www.jthink.net/jaikozforum/posts/list/650.page
http://www.jthink.net/jaikozforum/posts/list/873.page
Furthermore, if the user cares about the accuracy of ?hidden? tags (e.g., release country), then Autocorrect is much less accurate. Cluster albums can be helpful but has its limitations. Cluster will not correct a track that has been changed to a different release, nor will it unify the country of release for tracks with the same release ID. Cluster is best when you have a bonus track on an album and want to make sure all tracks reflect the bonus-track release, not the original release. Still, I always save before using Cluster, and review changes from Cluster before saving again.
http://www.jthink.net/jaikozforum/posts/list/841.page
For all these reasons, I prefer Manual correct tags. It is a huge time sink. Sadly, I feel compelled?.
Some tips/comments on Manual correct:
Sort by reverse load before launching Manual correct tags. If file names start with track numbers, then the last track will appear first. Since the difference between releases is often which bonus tracks are added at end, you can home in on the proper release more easily this way.
Watch the track length. Despite what you may have in the score settings, matches of wildly different length appear in the search results.
You can force match a track to a particular release by copying the release ID into the appropriate cell and then running Update tags from existing MB ID (make sure the MB unique ID cell is blank). This does not change release country or year, however.
Jaikoz does not remember the size & position of the manual correct result window between uses. I have to enlarge and drag it each time. It was a big bonus when columns and column order got saved?can we extend this to window settings too?
Personally, I don?t like that Jaikoz now rechecks songs that have a MB unique ID when Manual correct tags is run, but that is easy enough to get around via filters, sorting, etc.