Song Tagging and Metadata Blog

A SongKong customer asked me about a problem he was having with SongKong and JRiver Media Center this weekend. Although SongKong appeared to have updated his songs correctly when he looked at the songs in JRiver there was no change, restarting JRiver made no difference.

The problem is that JRiver (just like iTunes) build an index of all the metadata the first time the files are added to JRiver. The advantage of creating an index is speed, you can now browse your file metadata within JRiver very quickly without JRiver having to go back to the files themselves, it only has to go back to the file when you actually want to play the file.

The disadvantage is that if you modify the files with some other tool such as SongKong then JRiver does not know that anything has changed, you have to tell it.

Luckily this is easy enough, just use Edit:Select All to select all your songs within JRiver. Then select Tools:Library Tools:Update Library (from tags). This may take a few minutes depending on how large your JRiver library is. If your library is really large and you have only changed a few files then you can just select the modified files instead of all your files.

Introduction to uPnP AV for streaming Audio to Hifi

What Used to Happen

A few years ago digital music was usually confined to a single machine. For example you might download files from iTunes on your Apple mac and then play them on the same mac. You may well have an iPod, in which case you would copy your files onto to your iPod and they would exist as separate files.

At the same time Hi-end audio system typically consisted of a CD player, a DAC (Digital to analogue converter), Amplifier and Speakers.

There was little cross-over between the two

Better networking

Then we started to get hi-speed internet and wi-fi throughout are homes. Network attached storage (NAS) allowed us to just have one copy of our music , videos and photographs which could then be accessed by multiple devices.

Everything then started to become networked, not just computers but TVs, set-top boxes, Cd players and stereo systems. A way of having this wide range of different devices talk to each other was required. Microsoft had already developed Universal Plug and Play or UPnP for computers, gaming consoles and network devices such as routers. So it became the obvious choice as a standard for providing services between devices

What is uPnP AV

AV refers to Audio Visual - this takes UPnP and develops standard for audio and visual, i.e hi-fi systems, televisions, set-top boxes and the like. When people talk about uPnP they usually mean uPnP AV, on this blog we are interested in the A part.

Hi-end audio manufacturers such as Naim and Linn started to see the potential of using hard drives to store music rather than always playing from the original CD. But customers wanted to play music in multiple rooms this meant a way was required to stream music to players that could reside in different locations to where the music was stored, and they wanted to be able to control this from a single point, uPnP AV addresses these issues.

uPnP

uPnP AV has three main components

1. Media Server
2. Media Renderer
3. Media Controller

The media server streams music to media renderers, control of the media renderers is via the media controller.

Media Server

The media server is software that could be running on a computer, a NAS or a purpose built hi-fi server such as the Melco N1 or the Naim UnitiServe

The most popular media servers are Asset, MinimServer, Twonky and JRiver . The choice of media server is important because the Media controller can only see the metadata that the media server presents it with, and the media renderer can only play the files that the server provides.

We think MinimServer is by far the most flexible media server, offering a significantly more powerful approach to metadata then the others that let you harness all that metadata added by SongKong and Jaikoz, and we will explain how to do just that in a series of blog posts in the near future.

Media Renderer

The Media Renderer either plays the music itself or converts it so it can be played by something else further down the line. So the media renderer is usually a piece of hi-fi hardware such as the Linn - DS Series or the Naim NDS

Media Controller

The Media Controller is better known by the curious title of Control Point. Since customers want to be able to control their music where-ever they are in the house the focus of media controller development is definitely for iPads and Android based tablets. But control points are available for
the Mac and PC as well.

Control Points fall into three categories:

1. General purpose software that is designed to work with any compatible server and renderer such as PlugPlayer and SongBook

2. Hi-Fi Branded software that actually seems to work fine with any server, an example of this is Linn Kinsky

3. Hi-Fi software tied to particular hardware such as the Naim App

The choice of control point is not as critical as the other components, at the moment they all work in quite a similar way with not much to distinguish between them.

Jaikoz was originally released back in 2005, over ten years ago. It got me thinking where has the time gone, and what have we achieved ?

I have always tried to have original ideas rather than take other people ideas and I realized that actually there are quite a few features of Jaikoz and SongKong where I got their first. So at the risk of being self-indulgent here is a chronological list of some of these.

Jan 2005 Jaikoz is released with full spreadsheet interface, this meant instead of having to edit tracks one by one changes could be made rapidly.

June 2005 Jaikoz approach MusicBrainz to become the first commercial tagger to use MusicBrainz for tag matching and help setup the Tagger affiliate program. From the start we always stored the MusicBrainz Ids as well as the song metadata to allow users libraries to be fully MusicBrainz enabled.

June 2007 JThink open source our tagging library for reading and writing metadata to files for different audioformats called jaudiotagger. Jaudiotagger has been very popular and as well as being popular on Pc, Macs and Linux is now the No 1 tagging library for Android devices.

Sep 2008 Jaikoz is the first tagger to add support for Discogs matching when they opened up their metadata without restrictions.

Nov 2008 Jaikoz adds tools for submitting collection to MusicBrainz and seeding MusicBrainz releases.

Feb 2010 Jaikoz takes spreadsheet support further by allowing metadata to be exported to a spreadsheet, edited and then imported back into Jaikoz and hence the files. No other application can do this.

Nov 2011 SongKong released, the only totally automated tag editor with no options for manual editing. SongKong is the first application to use Acoustid for acoustic fingerprinting.

May 2012 SongKong adds a database based Undo facility that allows metadata changes to be reverted even after a reboot, this feature is unique to SongKong.

July 2014 We create the first combined MusicBrainz/Discogs database server called Albunack. This combines data between the two datasets, includes all metadata that could be used for tagging. It is optimized to provide much better performance than be achieved using MusicBrainz or Discogs directly. Shortly after Jaikoz and SongKong move to use this database.

Dec 2014 SongKong supports both release based matching and song only matching when the release cannot be accurately determined. Song only matching updates song details such as track title and track artist but not release metadata such as track no or album title. This is in contrast with other applications that either treat songs independently and take no care over their album metadata or applications that do consider songs as parts of albums but then prevent matches when the albums cannot be determined. Im not aware of any other application that can currently do this.

Feb 2015 SongKong is the first application to use AcousticBrainz to support Bpm, Key and Mood analysis. Jaikoz adds support shortly after.

Oct 2016 We release the first version of SongKong to solve many of the metadata issues people have when trying to identify Classical music. Solving issues with identification of Classical, Works, Movements, Composers, Performers and Artists.

Dec 2016 SongKong and Jaikoz become the only applications to support automatic updating of the new works and movement fields newly introduced within iTunes

March 2017 We team up with Melco to solve an issue for Naim UnitiServe customers, namely parsing the proprietary metadata used by Naim and embedding into the files themselves using standard metadata.Unique to SongKong for Melco.

In the Introduction to UPnP I explained that there are three components to UPnP AV, the media server, media controller (better known as control point) and media renderer. The metadata displayed in the control point is defined by what is provided to it by the media server rather than the control point itself. A control point cannot display data that the media server doesn't provide it with, although poor control points can limit the data displayed further.

I also said

We think MinimServer is by far the most flexible media server, offering a significantly more powerful approach to metadata then the others that let you harness all that metadata added by SongKong and Jaikoz, and we will explain how to do just that in a series of blog posts in the near future.

So in this post I going to explain the advantage of using MinimServer, in a nutshell MinimServer lets you serve up any metadata to your control point in the way you want. In contrast other servers tend to restrict to the standard Pop/Rock conventions of Artist/Album/Track/Genre/Year.

MinimServer was developed with Classical Music in mind, but there is nothing in the design that limits its use to Classical Music it is just that Classical Music simply better demonstrated the problems with other uPNP AV servers.

Jaikoz and SongKong have always provided one of the most comprehensive metadata coverage of any tagging application. But one frustration is that players such as iTunes, Winamp and MediaMonkey do not support many of the fields that are supported by Jaikoz and SongKong, the only way to make this data visible to users within these applications is using the data as part of the filename, and hope they will eventually improve their metadata field support.

But with MinimServer this data can be served to the customers control point, we just have to configure MinimServer correctly. So we can index songs by fields such as performers, groups, movements, orchestras, engineers, conductors, record labels. We can make use of the sort fields we have for not just artists and composers but other roles such as conductors and producers

In fact MinimServer has many configurable aspects, but lets start with the basics, in the next post we show a vanilla MinimServer installation and compare with another popular media server.

Today we release SongKong 4.7,this new release has some major improvements to the Fix Songs report, some new options for classical matching and fixes for a number of niggling issues.

View Metadata as Spreadsheet

An often requested feature is now here. The Fix Songs Report now has a new option called View Metadata as Spreadsheet, clicking on this will open a spreadsheet (.xlsx format) in your preferred spreadsheet application listing the complete set of metadata within the set of files processed by SongKong.

This shows the metadata as it currently is, if in Preview mode it shows the metadata as it would be if not in Preview mode.This spreadsheet view enables you to very quickly view multiple changes on multiple files.

Because SongKong can typically add a lot of different metadata we group the metadata into logical groupings over a number of sheets. The sheets are Basic, Release, Classical, People, Sort Fields, MusicBrainz & Discogs, MusicBrainz Work and AcousticBrainz - every sheet lists songs alphabetically by filename, the filename is in the first column of each sheet.

When SongKong has added a metadata for a field that did not have any metadata for that field it is shaded yellow, when the field did already have metadata but it has now changed then it is highlighted green, if a field is deleted it is highlighted red.

Classical Improvements

Before this release you could force all releases to be considered not classical but you could not force all releases to be considered. Also you could not allow SongKong to set the IsClassical field without applying Classical options. we have now replaced one option with the two below.

Identify Classical releases

SongKong uses various way to identify Classical tracks, if a song does appear to be a Classical piece then the IsClassical field is set. Usually it is best to let SongKong to decide if a release is classical or not by keeping the default option of Classical releases identified by SongKong . But some releases are difficult to assess. If you know all the releases you are matching in this run are Classical thenyou can set this option to All releases treated as Classical to force SongKong to treat all matched songs as classical and set the isClassical field. Alternatively if you consider none of the releases to be Classical then you can set this option to None, no releases treated as Classical to force SongKongto treat all the songs to none classical, clearing any values previously set for the IsClassical field.

Apply these options to releases identified as Classical

If this option is enabled then whenever a song is identified as Classical the options on this tab are applied to it, they are not applied to releases identified as non-Classical.

not_classical_release.txt

When SongKong decides if a release is classical or not there are still ambiguities regarding if a release is Classical or not since it is a little subjective.So we have also added an exception list . Adding the MusicBrainz Release Id of a release to the not_classical_release.txtfile ensures they are not treated as Classical releases.

Improvements to Song matching

Various improvements and fixes have been made to single song matching, single song matching occurs when an attempted match of a group songs failed to find a match. But then depending on customer options an attempt can be made to match individual songs.

Other Fixes and Improvements

We have made a number of minor improvements in many areas of SongKong, the full list of fixes can be found on the SongKong News page

We have added some important improvements tothe ability ofSongKong for Melcoto read All Music Guide data from the Naim proprietary WAV metadata format and embed the metadata in the files themselves so that can be read by non Naim applications.

Fix Songs Report - All Music Guide section

We now have a separate section in the Fix Song report called All Music Guide, this works in a similar way to Matched to MusicBrainz Release and Matched to Discogs Release

Selecting All Music Guide opens up an alphabetical index:

Then selecting a letter lists the Album Artist of releases that have been matched to All Music Guide metadata, together with a count of the number of songs that had been matched.

Selecting an Album Artist updates the right hand side of the report with the All Music Guide album(s) that have been matched

Selecting one of these albums lists more information about the album together with a list of all the songs that have been matched on the album and file that has been matched for each track.

Unlike Matched to MusicBrainz Release and Matched to Discogs Release SongKong is not trying to search for a possible match it is simply taking the amginfo.xml file for the files in the folder, parsing this file and embedding the metadata in the files themselves. There should be the same number of music files in the folder as on the release and therefore all the songs on the release should be matched.

Please note we talk about the All Music Guide in this article but when the Naim UnitiServe rips a CD and it cannot find a match to the All Music Guide it does a lookup from CDDBinstead and creates a cddbinfo.txt file. SongKong also parses and embeds these files as well, but they contain less metadata information than the All Music Guide.

Fix Songs Report - Song Change section

The song changes report now indicates for each song when they have been matched to the All Music Guide. In this example all songs have been matched to All Music Guide, MusicBrainz and Discogs !

Go forward to Minim Server:Part 2
Go back to Introduction to UPnP AV using MinimServer

Installation and Start

MinimServer is developed in Java meaning it can be made available for a large number of computer devices, from Macs and PCs to Network Attached Storage devices (NAS) such as Qnap or Synology and purpose built media servers such as the Melco N1.

There is a full list of installations available here but for this blog we are going to use a PC based installation. We are also going to do a brief comparison with another popular UPnP server called Asset just to show what difference it makes using a different media server. Since this is a music metadata blog we are currently just concentrating on the metadata features rather than control of the media renderer.

We already have Asset installed and have scanned a folder containing approximately two hundred songs in Flac format. Now with MinimServer installed, we double click it to start in system tray, the first time it starts it asks for a content directory, we then select the same two hundred songs folder to scan for music.

And when it is ready it can be seen in the system tray as a green icon containing of course a Minim

View with Control Point

Now lets view with the Linn Kinsky control point, although branded Linn it seems to work quite reliably with any media server. We have the same set of music files available to be served by both Asset and MinimServer.

Asset UPnP

Lets select Asset UPnP first

The first thing to notice is that when served by Asset various pseudo lists are available such as Top Artists and Top Albums and we have nice icons for different types.

We can browse by Artist,Album, Title, Composer and Genre

MinimServer UPnP

Now lets take a look at MinimServer

With MinimServer we don't have these additional list or icons ( I think icons would be a nice to have Simon).

However as well as browsing by Artist, Album, Composer and Genre like Asset we can additionally browse by Conductor and Orchestra. You can probably configure Asset to browse by conductor, I'm not so sure about Orchestra. But the key thing is this demonstrates the different focus of these media servers. MinimServer better supports Classical and is more album orientated. I think for this reason MinimServer doesn't include Title by default, but it is simple to use as we shall see in a later post.

MinimServer Configuration

Everything can be configured via MinimServers configuration panel, simply right click on MinimServer in the system tray and select Properties.

This will bring up the MinimServer interface

This is not the prettiest interface, nor is it very intuitive since to view the instructions for the various options you need to read the User Guide on the website.

But it is very effective, and not prescriptive, there is alot of flexibility available to you in this simple interface.The trick is to understand what you want the end result to be, how your songs are tagged and how to properly use the options to achieve what you want.

In the next post we show the first simple customization to make better use of metadata added by Jaikoz and SongKong.

Go forward to Part 3: Sorting Conductor by Sort Name
Go back to Part 1: Introduction

Continuing with the same dataset if we double click on the Composer index the composers are displayed alphabetically, but by FirstName rather than by LastName

It is more usual to organize people by the last name, for example consider a telephone directory, and this is the purpose of Sort Fields. A sort field isn't typically used for display but to sort the values for display.

MinimServer has a set of field names that is has mapped to logical fields that can be used for any audio format in the Tag mappings section of the User Guide. One of these is the ComposerSort field

MinimServer and Reverse Names

Whilst MinimServer can be configured to guess the sort name by reversing the values using the reverseName option this only works for the usual Western names. Sometimes an artist doesn't have a first name, and composers in non-latin script cannot be handled. It is preferable to use dedicated sort fields rather than guessing the sort value by looking at the non sort field.

SongKong and Jaikoz support for Sort Fields

When we look up a person in MusicBrainz they have a dedicated sort version of their name. This means the ComposerSort field always has a value and latin script sort names are provided for all people, even if their name is in non latin script such as Cyrillic or Chinese. For example the sort name of

Пётр Ильич Чайковский

is

Tchaikovsky, Pyotr Ilyich.

MusicBrainz data is peer reviewed and therefore generally reliable.

MinimServer Configuration

We configure MinimServer as follows:

We add any fields we need MinimServer to read from the files but not index, so we add ComposerSort, case doesn't actually matter we could have added as composersort or COMPOSERSORT.

Then in tagValue a special syntax can be used to tell MinimServer to sort a value by another value, so here we tell it to sort the Composer field by the ComposerSort field using the format

DisplayField.value.sort={SortField}

Tell MinimServer to rescan and now when we view with our control point we can see the composers are now sorted correctly by their last name.

Other Sort Fields

We can use this approach for any other of people such as conductors and performers in our metadata. Luckily SongKong and Jaikoz provides the required data, for example Conductor Sort is not a standard metadata field but SongKong and Jaikoz provide it so it can be used by MinimServer, we delve into sort fields more in the next post.

Go back toPart 2: Sorting Composer by Sort Name

In the last section we saw how to sort Composer by the ComposerSort field, lets now apply the same principle to the Conductor field. But if we look at the tag mapping section of MinimServer User Guide again we find that a ConductorSort field is not listed.

Some History

At one point the only computer music files commonly found were WAV format but at this time (when harddrives were more expensive and smaller) WAV didn't really support metadata and WAV files were too large to be that useful. But then MP3 was invented, originally with the ID3v1 metadata format and then shortly afterwards the ID3v2 metadata format.

The ID3v2 specification supported alot of fields, and was rather complicated. Music players and tagger software spent years trying to fully support ID3v2. As a result of this when other tag formats came along like Vorbis Comments (used by Flac and Ogg) and Atoms (used by Apple Lossless/Lossy) the software concentrated on ensuring they had support for fields equivalent to those in ID3.

ID3v23 had no sort fields, ID3v24 added ArtistSort , and then Apple added ComposerSort and AlbumArtistSort, and these became a defacto standard. But none of the big players has added ConductorSort and hence it is rare to see it in supported in any format or player.

SongKong Support

But the field does exist in SongKong andJaikoz, the information is freely available in MusicBrainz so it made sense to add this when tagging. All audio metadata formats do in fact support a standard way to add custom fields, its just that ConductorSort is not listed not as a standard field.

Finding Non standard Field Names

If we look at SongKong Field mapping table we can see that ConductorSort is listed slightly differently depending on the audio format.

TXXX:CONDUCTOR_SORT (MP3, WAV, AIF)
----:com.apple.iTunes:CONDUCTOR_SORT(MP4)
CONDUCTOR_SORT (Flac, Ogg, Wma)

The TXXX and ----:com.apple.iTunes are those formats standard way of creating custom fields and MinimServer understands this and can read such fields, all TXXX fields and all ----:com.apple.iTunes can be read by MinimServer. When you specify such fields in MinimServer you don't specify the TXXXor----:com.apple.iTunes parts.

Any field in Flac or Ogg format can also be read by MinimServer.

So we only need to look at the second part. For all the formats this is CONDUCTOR_SORT, this is great because it means that MinimServer can be configured the same whatever format you are using.

But if they were different and you had audio in different formats (e.g FLAC and WAV) then you would need to use tag aliasing.

Adding Conductor Sort to the Configuration

We take the same approach as we did with ComposerSort, but because we are using a non-standard field we use its actual tag name which for all formats is CONDUCTOR_SORT (note the underscore), case does not matter. Because we are adding another rule to the itemTags and tagValue we separate the rule with a comma.

And if we rescan and view in our control point we can see conductors are now sorted by their last name.

Orchestra Sort

We can apply the same mechanism for the Orchestra field. Orchestras are not individuals so the first name / last name rule does not apply. But the problems of sorting non-latin names and avoiding orchestras with 'The' at the start of the name all being grouped under 'T' do, so simply modify configuration as shown in the screenshot.

We can apply sort to the Artist and All Artists index as well, but there is much more to discuss with artists. Next post we change tack and show how to use SongKong with one of MinimServers killer features, Groups and Works.

Go back to Part 3:Sorting Conductor and other fields

Groups and Compositions

Minimserver has the concept of Groups and Compositions. In its purest senseGroups are simply a way to groups tracks within a particular album. Whereas Compositions are analogous to a Classical Work, all movements in the work should have the same value for Composition, the Composition field can then be indexed and browsed in the same way as artist or album.

The curious thing is that usuallytracks should have the same value for both (we will explain the differences later on). Lets return to groups, they are really most useful for Classical releases to indicatethat certain tracks are all movements of the same work.

Groups Example

Without using groups all tracks are listed in track order when you open an album

But with groups we see the groups representing a work when you open the album. This is much clearer, we can clearly see at a glance there are two different works, String Quarter in G minor Op 10 and String Quartet in F major. This makes sense because it is more common to play the complete work rather than just one movement of the work

but you can still double click on a group to see the movements contained within, and play individual tracks.

Configuring Groups with SongKong

So how can SongKong help?

MinimServer discusses a Group field, but this not a standard metadata field.

SongKong identifies Classical releases and the works and movements within those releases, the Works are stored in the Work field.

So all you have to do is configure MinimServer to read the Work field and use it for the Group field.

Add Work to itemTags so the Work metadata is read
IntagValueset Work as the default field to use for Group
Tell MinimServer to Rescan

Browsing by Works

So now we have tracks grouped by works within albums, but we cannot yet browse works from the top level. To do this simply move Work from the itemTags field to the indexTags field as follows:

and Rescan.

Now you can browse the Work index

If you open a Work Composition item, it then shows the Work Group

When Works and Groups are different

So we can see that usually it makes sense for Work and Group have the same value, but not always.

Sometimes movements are not listed contiguously within an album. MinimServer does not support non-contiguous Groups, but there is no such restriction on MinimServer Works.
For multi-level works such as an Opera, containing Acts and Scenes it may be preferable to have groups at the Act level and a single work representing the Opera itself

Configuring different fields for Work and Group

SongKong Works are currently analogous to MinimServerGroups.If you need different values for Work and Group SongKong can be configured to also write the Work value to the Grouping field by setting the Copy Work to Grouping field
to Yes

The Grouping field is mapped to the ContentGroupfield by MinimServer, except for Flac/Ogg. SongKong writes Flac/Ogg to the Grouping field which is not mapped by MinimServer. To use the Grouping field for Group, and Work field for the Workand have it work for all formats you can configure as follows:

Editing Values

At this point the values are the same for both, but you could usea tagger such as SongKong's big brother Jaikozto modify each value independently.

Future Improvements

We plan to add direct support for writing independent values for Work and Group that allow for the differences in use in MinimServer.

Next Time

Next time we look at artists

We are pleased to announce that a major new version of SongKong is available today

The main focus of this release is once again support for Classical Music tagging but there are other things as well

We now allow you specify almost any fields that you do not want Songkong to modify, and also any fields that you only want SongKong to modify if they are currently empty. Both of these options are on the Format tab
We have similar options for just Classical music so that you can configure SongKong to usually modify a field but not if it has been identified as a Classical song.
Added option for Classical to remove the composer from the AlbumArtist field, since the composer is not usually a performer.
Added option Shorten Work:Movement track titles to just Movement where possible - so your classical tracks just have the movement in the title.
Now Classical groups are categorised as Orchestra, Choir or Ensemble, previously the Orchestra field was used for all types.
Added isclassical as a field that can be used in rename masks
Added Ranking and Classical Catalog No fields, we already had Opus but these fields allow us to add other ways of cataloging classical releases
Fixes to allow encoding of filename on filesystems that do not support 16bit characters in filenames.

You can download the latest version from here

Forward to Part 5 - Classical Classification
Back to Part 3 - More Fields Required

One issue with Classical music is Classical composers were not the most imaginative when naming their works and it can be difficult to differentiate one work from another. Conversely the same single work can be available on many different releases but not consistently named by the record companies making it difficult to identify when you have different versions of the same work.

Some composers used the Opus catalog mechanism to organize their works, if used in conjunction with the composer this can be helpful for organizing your music.

Opus

Many Classical composers used the Opus (Op.) number. The concept was to catalogue each work as they were published, so the Opus provides a list of all the composers works in date order.

However some composers only bothered to catalog their best works, some only used it for certain types of works and some didn't use the system at all.

An Opus can contain multiple works. most famously Vivaldi’s “Four Seasons” was part of twelve violin concertos all written between 1723 to 1725 packaged as Opus 8 “Il cimento dell’armonia e dell’inventione”.

Sometimes works were only produced after the composers death.
For example Schubert’s last three piano sonatas, which were written in the last months of his life, but were not published until about ten years after his death are written Op. posth[umous].

Ranking

Whereas the Opus should identify all a composers works in date order, the ranking (No.) is used to catalog all works of a particular type

For example Beethoven wrote 9 symphonies, numbered 1 to 9 in chronological order, and 16 string quartets numbered 1 to 16.

Mozart wrote 40 symphonies, but these are numbered 1-41 because the Symphony No. 37 in G major, K. 444 was found to be composed by Michael Haydn and was removed from Mozart’s catalogue, so there can be gaps.

Opus with Multiple works

Remember an Opus can contain multiple works. With L. van Beethoven’s String Quartet No. 8 Opus 59 No. 2 in E minor, there are two numbers No. 2 and No. 8. The No. 8 refers to the ranking among his sixteen string quartets; but the No. 2 actually represent the sub-opus number (the second piece within the Op 59).

SongKong

SongKong intelligently extracts Opus and Ranking numbers into their own fields. Now they are in their own fields they can help you to search and organize your collection and they can also be used to construct new filenames for these classical compositions.

But other cataloging methods have been developed for certain composers and we will discuss that next

Forward to Part 5 - Classical Classification
Back to Part 3 - More Fields Required

Pleased to announce that SongKong 4.2 was released on October 18th 2016. This new release has a number of fixes and improvements

Add Composer to Album Artist

Classical music albums often contains the composer name as a part of the album title but not always. Different composers often named their works simply such as Symphony No 1 so without the composer name it can be difficult to distinguish one album from another. With this option enabled SongKong will add the composer(s) surnames to the start of the album title if the composer is not in the current title (retrieved from MusicBrainz/Discogs) to help with easier album identification.

Other Fixes

[SONGKONG-1051] - Ensure AlbumArtist always set even if album has no album artist
[SONGKONG-1053] - Add option to include composer in album title for Classical
[SONGKONG-1054] - Romanize non-latin artist names should apply to all people
[SONGKONG-1058] - Bundle Java with linux headless on website

[SONGKONG-1024] - When run songkong from cmdline log output from jaudiotagger goes to console

[SONGKONG-1052] - When Part derived from MusicBrainz has a leading space character

[SONGKONG-1055] - Discogs match/update ignoring isClassical Never Modify/Modify if Empty option

Pleased to announce a new major version of Jaikoz, this release contains many improvements, fixes and some significant infrastructure work.

Mood, BPM and Key Tagging

Jaikoz Pro users can now make use of AcousticBrainz - analysis has been done on millions of songs that are in the MusicBrainz database, when Jaikoz matches a song to MusicBrainz that has been acoustically analysed it can be update with the AcousticBrainz metadata, and this is all easily visible in the Mood tab of the Detail panel

DSF Support

DSD is a technology first used with Super Audio Cd's (SACD), it uses pulse-density modulation encoding a diffierence approach to other lossles formats like WAV and AIFF. The signal is stored as delta-sigma modulated digital audio; a sequence of single-bit values at a sampling rate of 2.8224 MHz (64 times the CD Audio sampling rate of 44.1 kHz, but only at ¹⁄₃₂₇₆₈ of its 16-bit resolution). DSF is a way of storing this format in files with metadata and is gaining in popularity on hi end audio systems.

Jaikoz fully supports DSF, including acoustic fingerprinting and all metadata fields.

New Never Modify/Modify If Empty Options

Previously most fields had their own option that allowed you to configure Jaikoz to always modify, modify if empty or never modify the field, and the option was repeated for MusicBrainz match and Discogs match. But it was very cumbersome if you suddenly wanted to change the option for a number of fields. We have now replaced by a list of fields that you can easily modify for each option.

Performers

Jaikoz now properly supports editing of the Performer and Involved Person fields, these fields allow you to add additional people involved in the song whatever their role, be it musician or art direction

Classical Tagging

Jaikoz now supports many new fields including many new Classical fields, these Classical fields are all visible on the Detail tab. Auto population of the new Classical fields can currently be done with SongKong but will be added to a future release of Jaikoz.

Full Change List

The full list of improvements can be found at http://www.jthink.net/jaikoz/jsp/news/start.jsp

Go forward to Part 6 - Works and Movements
Go back to Part 4 - Opus and Ranking

In the previous section we saw how that classical composers in centuries gone by often used the opus and ranking system to classify releases, we also saw they didn't do a very good job of it !

But in the case of the major composers after they died often a good soul came along and tried to catalog their works in a more consistent way. This is called classical classification and is an attempt to classify all of a composers works, it is completely separate from the record label catalogue given to a new CD release by the record company.

Usually catalogues are in chronological order but sometimes they are a thematic catalogue.
For example Johann Sebastian Bach, uses the BWV (Bach-Werke-Verzeichnis). The BWV catalog is thematic so a lower number does not mean it is an earlier composition, for example his cantatas are numbered BWV 1 to 244 and Violin concertos BWV 1046 to 1051.

There are no other composers whose work is catalogued using BMV numbers so it quite simple to extract them from the song/work title into the Classical Catalog field and that is what SongKong does. But things aren't so simple the following composers all use a H (Haydn uses Hob as well) prefix

Haydn
Hector Berlioz
Bohuslav Martinu
Gustav Holst

and of course there are song titles that contain H. without it signifying a classical catalogue no.

Luckily because the MusicBrainz database stores the composer separately SongKong only attempts to find a classical catalog no when it has identified the composer of the piece, and then it will search for the correct classical catalog pattern(s) for that composer. This ensures that when SongKong has extracted a classical catalog no it gets it right

This field offers another way to organize works by the major composers.

Go forward to Part 6 - Works and Movements
Go back to Part 4 - Opus and Ranking

There has been update to this regarding the Grouping fieldView Post

In October iTunes added some long overdue better support for classical releases in the form of new moment, movement no, total and work fields. This allows movements within a release to be shown together, and really helps for classical releases, especially when you have an album containing multiple works. This is exciting news for SongKong and Jaikoz and we are working on adding support for auto population of these fields in the next release.

However these new fields have caused a problem.

Why is my grouping column blank for newly imported files ?

MP3 and AIFF files use the ID3 metadata format for storing information about the music such as the artist or album name. ID3 did not have any fields (known as frames) defined for most of these new fields so Apple has defined some:

MVNM - Movement
MVIN - MovementNo/MovementTotal

But ID3 does already have a field that could possibly be used for storing the work, the TIT1 frame, this is how it is defined in the ID3 specification document.

The 'Content group description' frame is used if the sound belongs to a larger category of sounds/music. For example, classical music is often sorted in different musical sections (e.g. "Piano Concerto","Weather - Hurricane").

This definition is a bit open to interpretation so over the years the frame has been used by different applications in different ways, both as a work and as a way to group multiple tracks in other ways. For example in SongKong it can be used to store genres as a list of values because iTunes Genre can only see the first value, a problem if you want to categorize a song with multiple genres.

In fact until version iTunes 12.5.3 iTunes have always used this frame for the Grouping field. When you update to the version of iTunes your existing files will appear unchanged, but any grouping data that was stored in the TIT1 frame has now silently been moved to iTunes internal database only.

But when you import any new music that had the grouping field set you'll see it no longer shows in the Grouping field of iTunes but the Work field, and this also applies if you update any music outside of iTunes that is already in your iTunes library.

You can see this by going to the Songs view and enabling the Work column, and there are all the missing groups.

I created this test from SongKong very simply by
- Ensuring Update Genres and Save changes to iTunes was enabled on the Basic tab

- Setting Grouping to Always Replace Values on the Genres tab

- and running against a folder of MP3 files

So you can't use the Grouping field anymore, right ?

Well no, once you have added your songs to iTunes you can enter data into the Grouping field and it will persist. But it's no longer actually stored in the TIT1 field itself. Instead it is only added to iTunes internal database. But the problem with this is you can only fill in the Grouping field manually from within iTunes after the file has been imported in. Then once done to get that Grouping data out you would have to use iTunes AppleScript (OS X) or the COM API on Windows, this makes it very difficult to use your organize your songs consistently in both multiple places, such as using with both iTunes and Sonos.

Now for the really crazy part

MP4 files don't use ID3 for metadata they use atoms instead and the ©grp atom does continue to correctly map to the Grouping field if you import MP4 files.

This is because for MP4 Apple has defined the following new fields:

@mvn Movement
@mvi MovemntNo
@mvc MovemntTotal
@wrk Work

So there are separate fields for Grouping and Work !

Why didn't they define a separate field for Work for ID3 ?

That is the question, by adding new the new frames MVNM and MVIN frames they are already not strictly adhering to the specification anyway so the argument cannot be they did not want to break the standard

What does this mean for me ?

It means that if you use iTunes that when editing metadata for files outside of iTunes you need to consider different audio formats differently.

If you have a mix of Classical and non-Classical and you use AIFF or MP3 then it may be wise not to use the Grouping field for anything other than Works.

One question that requires more thought is when importing MP4 does it make sense to add tracks with the same Work as part of the same Grouping or not.

We will be adding options to SongKong and Jaikoz to make this as simple as we possibly can

Within iTunes it is probably best to decide on a format and stick to it, transcoding different audio formats to a single format.

In this previous post I explained the good news for Classical music fans that iTunes now has support for works and movements. But the bad news for Mp3 and Aif formats was that it was now using the TIT1 field to store the Work instead of the Grouping. This meant the Grouping field could only be set within iTunes and was stored in the database.

I have just installed iTunes v12.5.4.42 and this has changed.

iTunes has now added a GRP1 field for formats that use the ID3 metadata format such as Mp3 and Aif formats. The good news is that now you can set the Grouping and Work fields from outside iTunes by writing data to the GRP1 and TIT1 fields respectively.

But the bad news is that ever other player out there uses TIT1 for grouping so if grouping is important to you and you need to play your songs on iTunes and other players you have a problem.

Because if you set the Work field in iTunes, this will be seen as the Grouping field in other applications. If you set the Grouping field in iTunes this will have no affect in other applications because they don't support the new GRP1 field

It would have been infinitely more sensible for iTunes to leave the TIT1 field alone and define a new field to store the Work

We are pleased to announce the release of Jaikoz 9.1.0 and SongKong 4.3 today

Classical Feature Parity

We have made lots of improvements to classical matching in SongKong, but even better the whole classical matching functionality has now been added to Jaikoz. Classical matching now works the same in both applications and all the Classical music options that were in SongKong are now in Jaikoz as well.

Support for iTunes Work and Movement fields

Recently iTunes added support for Work, Movement and Movement No. SongKong already had support for deriving this information from classical releases but now it writes this information to the equivalent iTunes fields so you can now use SongKong to update these new columns for your classical releases. And even better news is that Jaikoz does this as well.

We will explain how this works in more detail in another blog post.

New Field Mappings

These it is quite customary to transcode music collections from lossless to lossy so that you have a high quality version of your music for your hi-fi system and smaller files for use on your iPod or car stereo.

We have noted that transcoders such as dbPoweramp do an excellant job of converting the audio, and a reasonable job of converting the metadata for the standard fields like album, artist and title. But they dont do such a good job of converting custom fields such as MusicBrainz Ids or the classical fields we have recently added such because they expect the fieldname to be exactly the same on all metadata formats. For MusicBrainz Ids this is not the case, i.e the MusicBrainzReleaseId field is called MusicBrainz Album Id for Mp3 and Aiff files but MUSICBRAINZ_ALBUMID for Flac and Ogg files.

We cannot change how MusicBrainz Ids are stored but we have revised the mappings for some of our recently added new classical fields in order to make it easier for metadata to be maintained when the files are transcoded.

This means in Jaikoz you many notice that some fields seem to be empty when you load them, just run Update Metadata from MusicBrainz to re-add the data to the newly mapped fields.

Acoustid Song Only Metadata

The Acoustid database is crowd-sourced and when songs are submitted to Acoustid they are submitted with their metadata. If this none of the songs submitted for a particular Acoustid fingerprint include a MusicBrainz Id then the Acoustid is not linked to MusicBrainz but the basic metadata - artist, album and title are available from Acoustid

Now when a song is matched to an Acoustid with no links to MusicBrainz metadata but with some basic metadata we use this metadata if our song doesn't already contain that information. In the SongChanges part of the report such songs are labelled (Acoustid Song Only)

This is great when your songs contained no metadata since the Acoustid database is larger than the MusicBrainz database. Also there are some songs that are in both Acoustid and MusicBrainz databases but not currently linked so adding this metadata increases the chance of your songs being matched to MusicBrainz by metadata (and Discogs).

When tracks contains featured artists

We have also added the When tracks contains featured artists option to Jaikoz this give you options when a track lists both a main artist and featured artists.

Go back to Part 4: Groups and Works

The Artist Index

Standard audio file metadata has two artist fields, artist and albumArtist. The albumArtist should be the same for all tracks on the release, the artist relates to the track and can therefore be different for each track.

By default MinimServer configures an All Artist index that indexes the (track) artist field, and a Artist index that contains just the albumArtist unless that is missing in which case the (track) artist is used instead.

Multiple Artists

Often a release or track is credited to multiple artists. Because many applications such as iTunes do not understand multiple values in these fields SongKong stores multiple people as one value in the artist and albumArtist field. But then stores the individual people in corresponding multi value fields artistsand albumArtists

MinimServer does understandmultiple values so its best to configure it to use these fields rather than the default.

We do this by mapping artists to artist and albumArtists to albumArtist, but we also have to map artist to nil and albumArtists to nil or we end up with the artist and albumartist values in the index as well. Use of nil is just a convention any value that is not used later on could be used.

We probably want to sort artists by their sort names as well. SongKong provides special sort fields that work with the multiple value artist fields called artists_sort andalbum_artists_sort so lets add these in as well using the same method we used for composer and conductor sortthis gives us:

Now when we browse Artists index we go from

Notice that in the first screenshot we have Dallas Symphony Orchestra, Edwardo Mata but in the second screenshot we have Dallas Symphony Orchestra on its own, Eduardo Mata is also listed alone (under M).

The advantage of this is you can now easily browse all the releases that Dallas Symphony Orchestra are involved in regardless of other artists involved.

What is a Classical Artist ?

We have seen how to deal with multiple artist credits, but what should the artist and album artist actually include: soloists, performers, orchestras, conductors or composers ?

We discuss this more in the next post.

Go back to Part 5:The Artist index

The Track Artist

In the context of Classical music the Track Artist only really exists as a concept in digital music. On a CD or record various different people can be involved in each track, they may be a soloist, part of an ensemble, orchestra or choir or the conductor and labelled as such, different tracks have different types of people involved. The Track Artist represents the most important people or groups involved in the performance.

The Track Artist is used in MinimServer in the All Artists index, it is only used in the Artists index when the album artist is missing.

There is no hard and fast rule about what the Track Artist should be. But SongKong lets you configure what fields to use using the MusicBrainz Track Artist option on the Classical tab. The default is Performers and Choir/Orchestra/Ensemble and Conductor - note the Composer is not included since they are not actually involved in the performance.

The conductor is involved in the performance, but since the conductor has their own standard conductor field you may decide not to add it to the track artist field.

There is also the option that just takes the first most important person/group Performers or Ensemble/Choir/Orchestra or Composer. This is quite useful if you just want to present a simpler value to MinimServer's All Artist index.

If you change your mind just change the option and rerun SongKong over your files to quickly update with the new values. Because the songs have already been identified SongKong does not have to rematch the files only update the metadata from the matched release.

SongKong does write all the individuals and groups to different fields as well so you can easily configure MinimServer to index performers, ensembles, choirs and orchestras as well. In fact, you may prefer not to include the All Artists index at all and just index the different types of people and group that can be included in the Track Artist field

Within MinimServer you would want to add the following

indexTags:Choir, Conductor, Ensemble, Orchestra, Performer_name:Performer

itemTags:choir_sort, conductor_sort,ensemble_sort, orchestra_sort, performer_name_sort

tagValue:Choir.value.sort={choir_sort}, Conductor.value.sort={conductor_sort},Ensemble.value.sort={ensemble_sort},Orchestra.value.sort={orchestra_sort},Performer_name.value.sort={performer_name_sort}

Note we use the Performer_name field instead of the Performer field since the Performer also includes instrument. But we rename the field to Performer using the OriginalName:NewName syntax so more user friendly. When configuring it's sort value in the tagValue field we have to refer it to by its original name (Performer_Name) rather than its new name (Performer).

Next time we look at the Album Artist