This partially contradicts itself, but I could see how this probably helps in your case
To understand what I mean: If you do a lookup Picard uses the existing tags as search terms. Especially it uses title , artist, album, tracknumber and the track length (it uses a few more, but those are the ones that commonly are filled in even for badly tagged files). If there would really be all tags removed a search would yield no proper results at all.
Now if you have a collection of random songs from compilation sources with probably wrong or misleading album names and track numbers this would affect the matching to the original release albums. But if a file would have absolutely zero tags Picard falls back to guess from filename. And this is probably what is happening in your case: Tags guessed from file names probably work better than the existing tags.
Instead of your rule of âAlways delete all tags before you startâ I would say âAlways remove all bogus tags before you startâ. E.g. if you know the album and tracknumber tags in your files are all bogus remove those, which will make Picard to search only for artist and title. Together with the slider settings you have this can improve the hit rate for albums.
This also clearly shows the strength and weaknesses of lookup vs scan. If you have somewhat ok tags lookup can perform very well, because the tags can indicate quite clearly what release this should be. Even more so if you tag entire albums.
But if your tags are missing or just âAlbum 1â, âTrack 1â by âThe Artistâ they are useless for a lookup, searching for these tags will not give you meaningful results. This is when you use scan, which looks at the actual audio content. This works really well in identifying the proper recording, but as it has no other information then the audio content it canât really decide which release to choose from. Actually in the end it still uses metadata comparison for finding the best fit if there are multiple results, but overall lookup works usually better if the tags are not too bad.
Ok, I should have said âbut keep Title and Artistâ, which is what I do. The rest will only screw things up when you want to match songs taken from compilations (sometimes quite dubious).
With a little over 300 songs (there are 305 Iâve kept) from 1950 to 1979 I guess the success rate for Lookup and Scan, combined was around 30%. Another 20% if you checked versions and changed to the correct one. Less than 1% false positives (I think I had 1 or 2 that Lookup got the wrong artist on). 20-30% were easy to fix with Lookup in Browser. The rest, around 25% were harder to get. About 5% were really hard, as in, non existent, and I had to add them. 2 or 3 I just deleted because 1. They werenât really that interesting, and 2. I actually didnât find much info at all, on the net (Milord by a Dutch artist was one). I also deleted all Betales songs sung by other artists (I have all the albums on CD and prefer the originals over more or less successful covers).
The group of Billboard yealy chart topper songs that is being worked on in this thread may be strongly related to the group of songs being worked on in thread: âScans producing totally incorrect resultsâ
Yeah Iâve been removing all tags except title and artist so I guess for me and my tastes which never got digital releases until years later the fingerprint thing ainât gonna find the vinyl releases with the proper dates etc⌠Oh well, câest la vie.
With much of my vinyl rips Iâve already tagged the main details including cat no using iTunes⌠I wish I could import all that into the add release forms in clusters⌠rather than one song at a time.
