Metadata, fixing things

January 3, 2016
 by Paul McGowan

Metadata is digital audio’s underlying information of what tracks are what, the name of the performing artist, band, cover art, liner notes, sample rate, filter type, release year, track number, etc. In short, metadata has all the information we need except one: the actual music itself. Without metadata we don’t know much of anything about a track of music.

Few of our libraries are perfect. I have spent hours upon hours curating mine–grooming it to perfection. And yet, there’s still unknown tracks and artists within my group of tunes. Or maybe a few duplicates, or some without cover art.

I remember one such case. I received a demo disc from someone at a tradeshow. It had all sorts of great music; some I recognized, many others I did not. One particular track I swore must have been by Elvis Costello. No, I reasoned, it can’t be Costello because it sounds too much like Buddy Holly, yet…the recording quality sounded modern, much better than anything I associated with Holly. Turns out I was incorrect. It was Buddy Holly off an obscure album I had never heard of called Buddy Holly, From the original master tapes.

How is it that I found out the name? I ran it through a program that fixed much of my library. This amazing program not only identified 100% of every unidentified track I owned, it also found duplicates, fixed names, years, and provided better cover art than I had. It was the best $32 I had spent. This was some time ago and now that program has been updated and improved and it is still the best $32 a person with a library can spend.

The program is called SongKong. Yes, a bit goofy of a name and the website’s not much to write home about, but it is a rather brilliant piece of work. And most important, it works.

How, you might ask, does it figure out a track of music without any metadata? The process is rather amazing. It’s known as acoustic fingerprinting. Like figuring out the identity of a criminal in a who-dunnit, acoustic fingerprinting makes a copy of the music and plays it into a database to locate a match. Once found, all the metadata is added back to the track and voila! We have the name, the artist, everything we might wish.

There are other programs like tuneup, MusicBrainz Picard – which also uses acoustic fingerprinting – or MP3 Tag, but I find all the others lacking in one way or another. SongKong gets it right, works on any platform, and is the tagging fixer upper of choice.

Subscribe to Paul's Posts

21 comments on “Metadata, fixing things”

  1. iTunes Metadata:
    Adding your own album art (e.g. from digitized LP’s) into iTunes if you’ve imported WAV files requires converting the files to Apple Lossless (ALAC).

    Thanks for the NAS articles, will check out SongKong!

  2. Great info ! Paul, does SongKong differentiate between original and remastered versions of a particular album ? It’s probably to much to ask for, but it sure would be nice to know who created a remastered version.

    1. Its likely that both versions are in the database and often but not always the acoustic fingerprints for a remastered track will be different to the original enabling a match to the correct version of the album. SongKong can also make use of existing metadata and the files locations (i.e what other files exist in the same folder) as well as acoustic fingerprints to find the best match.

    1. Actually DSF support will be added soon. Ive already added read support in the underlying library that SongKong uses and its quite straightforward to extend this to write support since DSF uses ID3 for its metadata just like MP3s and AIFF.

  3. The potential discrepancy between the various database providers is something to ponder although I’m guessing a lot of normalization takes places with crawlers.

    SongKong uses the open source MusicBrainz whereas iTunes likely uses the proprietary GraceNote or maybe Apple has created their own DB.

    Regarding cover art, unfortunately, I still find myself laboriously scanning CD covers for some classical titles where only a 200×200 image exits online. Not a big deal. I applaud the people at Chandos for maintaining a complete offering of all cover art in large formats.

  4. Great info – thanks! Another option to identify music is of course to use Shazam on your phone. It identifies songs you listen to and I use it frequently at trade shows but also when I have files without tags.

    1. I was fooling around with it just now. There are a couple of options for how it handles the integration, but in demo mode it doesn’t do anything — just previews what it could do — so I recommend downloading it and giving it a try. From what I can tell, it modifies the metadata on the files themselves, but keeps a database in order to undo previous modifications should you decide you made a mistake in changing the metadata.

      1. Correct, in trial mode it does everything except actually modifying the files, but you can see exactly what was matched and what changes would have been made in the report.

        Modifications are made to the files themselves, but it also stores the changes in an internal database to allow all changes to be undone at a later date, and (if configured) it also informs iTunes of changes and renames made to songs under iTunes control.

  5. Interesting, and I will try the SongKong demo soon. Not sure whether I’d want to apply it to my whole library, but it would be interesting to see what discrepancies it finds in the parts that are mostly curated, and how it can help organize the third that aren’t.

    I’m especially curious whether it can find art for singles. And, of course, whether it can correctly ID original albums vs. compilations…

  6. I purchased SongKong and ran it on my main music library. I don’t use iTunes so I didn’t test that part. I ran into two issues that SomgKong discovered but were not its fault. One, the library on my NAS was littered with lots of trash from Windows and Apple systems, including .AppleDouble and [email protected] directories. I had to SSH into my NAS to clean these up to prevent them from generating errors in SongKong. Second, for some reason, a number of my files had a read-only attribute set for the group and this prevented SongKong from making any changes. Again, I had to SSH into my NAS to fix this before SongKong could finish making changes.

    The only downside I have seen is with Classical music where SongKong will change the artist from its current value which is usually Conducter/Orchestra to Composer. Also, it does not update any information for the Orchestra attribute (I don’t know if this is a standard attribute but I use it always in JRiver). It also wipes out any soloists who might have been included in the Artist field originally. I don’t know if there is a option to correct this or to have SongKong skip Genre=Classical or something like that. Otherwise, I will have to do a lot of manual tag editing after SongKong applies its changes to my classical material.

    1. Thankyou kslano for your feedback, actually I think the first issue with dot files is something SongKong could deal with better, if you could let support at jthink dot net have your support files using Help:Create SupportFiles that would be a great help.

      Classical is difficult, typically composer (but not always) is set as the artist in the databases used by SongKong (MusicBrainz and Discogs) which is why SongKong uses it as the artist. We don’t currently use Orchestra, in ID3 (used by MP3s, AIFF, WAV) this is a user defined field called ‘TXXX:orchestra’, but I think that possibly only JRiver currently uses this. Different behaviour for releases known to be classical is an interesting idea worthy of some further consideration and I’ll raise an issue but it is not currently possible in SongKong.

      But you may not realize you can undo changes because SongKong stores them in a database, so if you decide that the changes makes to some albums is worse than the original file you can use Undo Changes and select the particular folder(s) you want to undo changes for, this can be applied to the current location or the original location (if they have moved).

      1. An update. I ran SongKong against my library and after spending every night this week cleaning up the damage it did, I would not recommend it for a library of any size. Among the problems I’ve had to clean up is around 300 files that could no longer be scanned by JRiver because the names created by SK were too long. As ijabz mention, SK does have an undo button but doing that an album at a time is tedious, not to mention the the additional time to clean all kinds of tmp and old files it leaves around, as well as lots of stuff in the recycle bin. The next problem is that it overwrites perfectly good (and correct) artwork with all kinds of garbage. Things like back CD covers, foreign covers, and other funky images. After getting all of that fixed, now I am finding all sorts of new albums in my folder that have only one or two songs in them. Now I have to look into how that happened but I suspect SK is mismatching songs from one album with another and splitting up my albums. It will take a lot of additional time to figure out what to undo to fix this, so I have my work cutout for me this weekend just trying to get my folder back to where it was.

        So a strong word of caution to anyone who wants to try this program, make a complete backup of your directory before using so that you can restore back without going through what I’ve been doing. While I agree with Paul that MusicBrainz and MP3Tag are harder to use, they never did this kind of damage to my library.

        1. Hi ksalno, really sorry to hear you are having difficulties. Have you emailed me about this because I do always respond to emails, and can usually resolve problems easier with the help of a customers support files. With your support files I can definitently work out the split albums problem, if the files aready existed one folder per album it should not happen. Support files are large but if you could take the trouble to make them available to me it would really help me to help you and improve the products for others.

          Regarding file lengths, I assume you mean Windows 255 characters limit, SongKong has code so that it abides by this limit so if you support files would help me diagnose this.

          Regarding Artwork, artwork from MusicBrainz is always front cover, artwork from Discogs is usually front cover but sometimes a Discogs release just has cover art without marking one as the primary artwork. This is usually front cover art, but not always. Perhaps I should have an option to ignore these. Updating artwork is not enabled by default, and if it by default it only replaces artwork if ‘better quality’, however better quality is based on dimensions rather than liklihood of being front cover so I need to look at that.

          Regarding Undo feature it doesn’t have to be done one folder at a time, you can select multiple folders and also folders within folders are processed.

  7. I reviewed ksalnos support files and thought it would be worth sharing the findings for the benefit of other users. Ksalno ran SongKong against a NAS drive mounted as Z:\ on his Windows system. Unfortunately there was a folder called Z:\Recycle Bin that was not actually a Windows Recycle Bin just a standard folder (at least as Windows was concerned) so SongKong was run against that in addition to the intended music folders. All the folders containing one or two songs are simply songs that were matched in the Recycle Bin and moved to new music folders based on their metadata, none of the files in the original music folders were actually split up. A problem such as this can simply be undone in a single step using Undo Changes, selecting Z:\Recycle Bin and setting ‘Find Songs’ to ‘that were originally in the selected locations’. To safeguard against this unusual case I could add a check to SongKong to ignore files that look like recycle bins based on a filename check.

    The 300 files that were too long for JRiver were within the Windows filesystem limits, the limitation was within JRiver itself. But I’ve posted a solution on the Jthink blog that lets you restrict the file length to your chosen length.

    According to the report, 95% of songs were matched, 91% with artwork. Artwork is not sent with the support files so I’m not sure how many had problem images but I suspect it is a small minority, although I recognise there maybe there are some things I can do to improve this.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram