#musicbrainz

New post: blog: Server update, 2014-05-26 <http://blog.musicbrainz.org/2014/05/26/server-u...;

the CAA is out to get me. looking at the stats page, the red line stopped being a factor when i went to bed. as soon as I woke up, it was back. silly servers plotting against me…

well, yeah. but I can think I’m special for a minute, right? heh.

hi

MBach: hello. did you have a question?

true

?

what is the best way to use the musicbrainz's webservice

in order to get all cover art

from a list like "artist -> releases"

all the cover art for a particular album, or a particular artist, or?

to mimize the number of calls to the webservice, from a particular artist

and let's say, 5 albums on 10

do you want just the front covers or all images?

(i am making an audio player by the way))

do you want all versions of releases, or just a representative cover?

i assume that the representative cover is enough

MBach: have you read http://wiki.musicbrainz.org/CAA ?

yep i have discovered it a few hours ago

i have started to play with the api

MBach: ok, in that case you'd use a release group browse request by artist mbid to get the release group mbids, then get the cover art image by release group mbid

the release group browse request is paginated, and can get up to (I think) 50 or 100 release groups per request

if you have a list of specific releases that you want cover art for, it's a bit more tricky

hmm

that's what i thought

but in that case, you should already have release or release group mbids, and no queries are needed

so not actually more tricky after all :)

(if you don't already have mbids, but are looking up e.g. by name, then you're gonna have trouble)

that's right :D

in that case, you'd be doing a search request for each pair of (artist,release group) to attempt to get an mbid, then using that mbid to get cover art.

ideally, you should store the mbid for things once you've done the lookup once so that you don't have to do it again... we have standard metadata tags to store this info in files, of course.

i read in the wiki that I shouldn't not do more than 1 call per second

yes, 1 call per second per ip address is enforced by the rate limiter on the api

(it's actually 10 calls per 11 seconds, but staying at the 1/1 rate will avoid issues completely)

er, 11 calls per 10 seconds?

whatever. 1/1 is safe :)

most of the people like me who are avid musicbrainz users will have files where the metadata tags were saved by musicbrainz picard, meaning that the MBID tags are in the file. In this case, you can get coverart without needing to do any api requests at all.

but in general I suppose that doesn't work :)

this is what i have done so far : http://imgur.com/8uwSiK7

sorry kepstin I always make my softwares :D

so you'll want to make fairly heavy use of caching - e.g. look up the mbid once when adding an album and save the cover image locally, so you don't have to re-query all the time.

yep, I want to create a small plugin which grabs missing pictures

you're writing your own player, and running foobar2000 in the mean time? :)

yep, because mine isn't finished =)

and foobar rocks

yours looks more finished than mine ;)

hmm.

i could help for french :D

but not for japanese

e.g. I see you have some Susumu Hiarasawa; for her 2012 release, we only have the japanese titles: http://musicbrainz.org/release-group/54e6c2d5-8...

although we don't have cover art for that anyways :/

so, let's back to work for just one pair (artist,album) right now

one little step at a time

http://wiki.musicbrainz.org/Development/XML_Web... is the api that you'll be using

you'll be doing a search for release groups, attempting to match the artist and album title fields

at work I'm more a Python guy, so dealing with HTTP Request and XML is a bit tricky

but thanks to Qt

(it's so awesome)

picard - the musicbrainz tagger software - uses PyQT :)

anyways, the type of query that you'd be doing would look like http://musicbrainz.org/ws/2/release-group/?quer...

note that you have to do some escaping of the artist and release fields there, to fit the lucene search syntax. Just a few special characters to escape, I think.

https://lucene.apache.org/core/4_3_0/queryparse... lists the ones to escape

the search will almost always give you back multiple responses, and you get a score with each one; based on the score and maybe some local comparisons, you can probably decide if there's a good enough match to use for cover art.

yep

sometimes using google nice too ;)

just type in what you want

and you have the right cover !

yeah, but that's not super useful to programattically find cover art :)

anyways, from that musicbrainz search, if you pick a release group, you just throw the ID into a cover art URL of the format http://coverartarchive.org/release-group/d77f1e... and download the image

er, and apparently that one doesn't work

yep, then you just hit the problem that our cover art coverage isn't superb :)

arf

so google was a not so bad idea, finally

aside from the fact that programatically doing google searches is a real pain, they have pretty heavy api usage limits iirc.

and if google doesn't find any good images or doesn't have a matching image, you don't really have anyway to tell.

i am making open source software, so I'd rather use open source from top to bottom

like TagLib

as an example that actually works, the search http://musicbrainz.org/ws/2/release-group/?quer... leads to the image http://coverartarchive.org/release-group/aa997e...

(and if you open that image, you'll note that it redirects a couple times before hitting the real location)

nice