Advanced MetaData Handling / Tagging Tool

[EliC]

I have pulled this out of my other thread as I think it is important and powerful enough of a feature to draw users and allow for much better functionality.

Original thread: http://forum.dbpoweramp.com/showthread.php?t=11922

Below is a quick mock up of a proposed worksheet to work with multiple meta data services at once and choose the best meta-data for tags. I choose the following album because its a good example of the differences between the databases and the fact that none of them are perfect. The meta-data in the final column most closely represents the data from the album. AMG is actually very good and only differs in this case a bit for meta-data style. Though there are plenty of examples were AMG is just plain wrong.



Meta-Data from AMG or MusicBrainz is given 2x the weight of FreeDB data since they are more carefully (but perfectly groomed). If multiple sources agree on the meta-data is is automatically moved to the final column. If sources do not agree then one of the sources must be selected or the data can be typed manually. Green next to the meta-data means the sources are in agreement. Red means there is a consensus the meta-data is wrong. Greyed out meta-data has been manually selected. If there is only one source for the meta-data it is trusted.

Hopefully we would also then be able to submit the corrected meta-data to all the sources.

-AMG: All users of this advanced tagger would be required to have AMG. Spoon has a contract to fullfill with them and they have good metadata. The solution, sell the tool as a separate addon only available with an AMG subscribtion.

-GetDigitalData: unlimited access for $5/year, with no commitment for numbers from illustrate.
# Artist Field
# Album Field
# Track Field
# Album Artist
# Genre (narrowed to approx. 32 specific genre fields)
# Cover Art (400x400 images)(as opposed to 250x250 from AMG)
# Year published
# Publisher
# Composer (classical genre)
# Conductor (classical genre)
# Compilation Information (Individual track artists for compilation albums)
# UPC & ISRC if required

-MusicBrainz: seems to be free, as would incorporation of their PUID generator

-Tracktype: free

-freeDB: free

-MusicDNS official support would likely not be free, but it would be easy enough to add a framework so end users could access it for free

In addition to the required AMG subscription I would be willing to pay for this addon. Hopefully it would function both with the CD ripper and as a library tagger.

Anyone else have any ideas/feedback?

[Teknojnky]

Re: Advanced MetaData Handling / Tagging Tool

Some good ideas there, I'd definately like to see more variety of metadata sources. Musicbrainz is my top source, albeit currently without respective album art.

[EliC]

Re: Advanced MetaData Handling / Tagging Tool

I asked musicbrainz about both album art and lyrics, but as you know both of these are "protected" by copyright. They do not feel they are big enough to license these metadata sources. As noted, GD3 seems to provide a better album art source then AMG.

Personally I prefer AMGs method of pulling featured artist out and listing them as multiple artists. I believe MB will be moving in this direction with their next major update.

I forgot to note it above, but AMG also supports acoustic fingerprinting (like MB).

Also, its not just about having any one source available over the others. None of them are perfect unfortunately. Its about being able to combine the power of all of them and pull them together easily and with the click of a button update any field to the correct value. Of course values in the final column would be text editable. Also, to narrow the size of the sources window a scroll bar along the bottom would be helpful.

[bhoar]

Re: Advanced MetaData Handling / Tagging Tool

One concern is this: what happens when the various metadata services return two or more matching albums? Can the worksheet be reworked to handle that situation in a plainly understandable manner?

This triggers the neurons to explain a tool that I'd love to have: a ripper and tagger combo that stores all of the rip-time lookup query data in a ready to re-lookup format (freedb/cddb TOC data or hash, as well as the others that use more/different query data) as well as all of the results pulled so far. That is, special tags or files would contain meta-metadata.

This is so that you can perform corrective lookups *after* the ripping process is through without having to rerip, manually search, or go back to find which CD sourced the normal-tag-free tracks. I'm of the school of thought that manual tag-correction during the ripping process drives me crazy (a bit ADHD here, can't really stand to come back in the room after being distracted for an hour to find the ripper stuck on the second disc).

Another benefit would be that it allows for more easily clarifying the results of CD singles (which almost always end up returning multiple matches). Along the lines of Eli's recommendations, you could also set an uncertainty threshold for his approach where: if the weighing of the various bits of data shows very close final numbers (no "clear winner), auto-tagging is not performed and the user can choose at a later time to finalize which data goes into the tags manually.

This would also be of help on Spoon's "Industrial Ripper" project, in that it would put all of the manual steps into one small part of the workday, instead of having to babysit the ripping process. It would be one time-saving step beyond the typical way this is done now (I think): difficult to tag discs are set to automatically reject and then they are later re-ripped, with supervision for corrective actions, after the full batch is done.

-brendan

[EliC]

Re: Advanced MetaData Handling / Tagging Tool

One concern is this: what happens when the various metadata services return two or more matching albums? Can the worksheet be reworked to handle that situation in a plainly understandable manner?

-brendan

I think the easiest way to incorporate this in the gui is to simply display all of the matches. If they dont match what other services are reporting they wont be used automatically. You could add an X next to the service at the top of each column and quickly delete the column and its contents if its completely wrong.

[Paul_B]

Re: Advanced MetaData Handling / Tagging Tool

I use Mp3tag after ripping with CD Ripper as this allows me a lot of control and access to sources such as Discogs and Amazon

[bhoar]

Re: Advanced MetaData Handling / Tagging Tool

I use Mp3tag after ripping with CD Ripper as this allows me a lot of control and access to sources such as Discogs and Amazon

Hmm, I saw mp3tag's features that seem similar but had very little luck with the post-rip freedb lookups. The failed IDs from the lookups didn't seem to find matches. When using the "determine from selected files" method, it would only find a match 1/3 of the time. In this case, the mp3tag-generated ID would be similar to, but not quite a match of, the discid that failed to match originally.

I guessed the issue is that one or more tools might not be generating perfectly compliant disc IDs. Maybe what I need is the ability to run an executable before each rip starts so I can use cdrdao to create the definitive cddb/freedb ID index and store it with the ripped files.

The discogs menu item always gives me a runtime error, so I haven't used it. However, that's a reminder for me to email them the error message...thanks.

-brendan

[EliC]

Re: Advanced MetaData Handling / Tagging Tool

It would be very easy for the original freedb disc id to be written to the tag. AMG also has a disc ID which should be written to the tag as well.

My experience with mp3tag has been that it is VERY labor intensive. The idea here is to be able to quickly and easily pull together as many resources as possible and utilize them as best as possible. I want to quite spending time tagging, moving, and renaming my music and just listen to it.

[Spoon]

Re: Advanced MetaData Handling / Tagging Tool

Instead of IDs you are best embedding the CD TOC into the tag.

[EliC]

Re: Advanced MetaData Handling / Tagging Tool

Is the CD TOC all you need?

[Spoon]

Re: Advanced MetaData Handling / Tagging Tool

That is all you have when you are ripping from a cd.

[bhoar]

Re: Advanced MetaData Handling / Tagging Tool

That is all you have when you are ripping from a cd.

Spoon, if you have time/interest would you let me know if the following statements are correct:

1. The standard cddb/freedb discid is a hash calculated directly from the information gathered from a CD's TOC.

2. Some other metadata lookup services use the same discID above.

3. The rest of the lookup services generate a different discID (again always generated from the CD's TOC) or use the TOC data directly. One reason they do this to get around weaknesses in the hashing function used in the original cddb/freedb system, which allows for too many collisions when fed differing TOC data, especially for CDs with low #s of tracks (e.g. singles or some classical CDs).

4. Accuraterip uses both TOC data and track data (the audio) with no fuzzy matching: original samples and strict sample alignment is required (that's the entire point).

5. Music *fingerprinting* services also use (or only use) track data and some can fuzzy match track data that has been lossily compressed (the track data does not need to be pristine). Sometimes this is bundled with their service that would fall into categories 2 and 3 above.

-brendan

[Spoon]

Re: Advanced MetaData Handling / Tagging Tool

1.Yes
2.Yes
3.Yes
4.Yes
5.Yes

[EliC]

Re: Advanced MetaData Handling / Tagging Tool

5. Music *fingerprinting* services also use (or only use) track data and some can fuzzy match track data that has been lossily compressed (the track data does not need to be pristine). Sometimes this is bundled with their service that would fall into categories 2 and 3 above.

What are you referring to as "track data" here? Most fingerprinting services are using an acoustic analysis algorithm to identify the song, unfortunately this alone is not enough to identify an album of origin. Most likely we are saying the same thing but it wasn't clear to me. As far as I know AMG, MusicBrainz, and Gracenote all offer fingerprinting services. I'm not aware of anyothers.

On a different note I have wondered if adding track and album gain values to an acoustic fingerprint service would allow for better album association.

[Dreadlord]

Re: Advanced MetaData Handling / Tagging Tool

MusicIP also offers acoustic fingerprinting....