#musicbrainz

Is mb involved?

Eh?

gioele: you have to speak louder

FEW DAYS AGO I READ THIS: http://blog.last.fm/2007/08/29/audio-fingerprin... IS MB INVOLVED IN IT?

DO YOU COPY?

8)

roger roger

I think Muz's hearing aid is malfunctioning though

Muz, can you turn for a sec?

turn away

No, you just worded it like a fucking moron

See, two can play at the asshole game ;)

Muz, please. Hey, look there are some lovely moose around the corner :)

meese?

None of the fingerprinting stuff is to do with MB, we developed it all ourselves

mooses?

Further down the line we may tie Fingerprint data + our metadata to correct names, based on an import of MB data to our database

But that's further down the line

And the plural of moose is moose

because I fear that "wisdom of the crowd" does not apply to music tags

Depending on the music, it does. Given there are 24153 plays of Britney Spears - Toxic and a lot lot less of Britney Spears - Txoic or any variant thereof

At least it did not apply to my friends' collections before I introduced them to picard

well it could work for spelling errors.

It will work for anything as long as the audio is correct

That's the whole point of audio fingerprinting

You could have no tags at all, and it should be able to match up the track to the tags correctly

recognizing stuff will work for sure. I wonder if it will work good enough to create a collection of good quality track titles (U vs You, uppercase vs lowercase...)

muz: just not to the correct release, but that's another topic

outsidecontex1, actually, it could

gioele, well like I said, there'll be match ups of our data to MB moderated data

And also our own user moderated stuff, like we used to have 3 years ago

Muz: great, this should solve almost all the problems

outsidecontex1, the fingerprinter records album metadata, and seeing as tracks vary across releases in time, track number, album name, and in some cases the actual audio, this should work well

muz: but if you have a track with no tags at all, only his fingerprint, and the audio was extactly the same on two releases there is no way to match it correctly. or do i miss something?

outsidecontex1, well the data wouldn't be accepted as definitive based on the fingerprints alone, which is where the user moderated side of stuff comes in

Also, with time, given other people fingerprinting the same audio with more complete tags

outsidecontex1: we have the same prob with PUID-based tagging in picard

The first step is just to gather as much fingerprints as possible before committing any actual data changes based off of them

muz: ok, i get it. i think we were talking from slightly different things, i was only considering the matching against existing data not the supplying of this data.

gioele: i know, it's of course a general problem. i don't think there is a real solution for it-

outsidecontex1: context :)

:-)

I call it the "Tuttapposto" problem

I have an album Tuttapposto in its 16-track version

PUID search insist on selecting the 15-track version

so one track is always excluded by the scan/match mechanism

if picard could use some of the context it has ("look there are 16 tracks to be scanned....") the problem would go away

gioele: yes. i think there is even an open ticket requesting that. would be really nice :-)

On the other hand such a context-sensitive scan would cause problems with uses of picard like "here is my huge collection of file: tag it"

Muz: thanks for the explanation

outsidecontex1 is now known as outsidecontext

Cat?

Oh well then.

warp: Not, at the moment. CatCat: What'd you like it to look like?

faux: It had better by shiny

I could add pictures of ponies?

Maybe make it bright pink?

perhaps lavender

Heh. No. :P A serious question, should /all/ trans*ated track-listings have the same cdtoc, or does that not happen?

FauxFaux: I'd say only if it's a pseudorelease should it be 100%

theoretically you could have a official release in Russia and a official one in the US, same release, different language/script, and perhaps not the same toc

Hmm, yeah, I could expect that to happen, as much as it'd make me want to stab people. :p

Bright pink :D

dumme freso, i stepped outside for liek a sec and ofcourse he dropped in to ask of me the bit I was out, ugh, satay more thna a sec next time :P

muz: pictures!

FauxFaux: huh?

wat look like who?

canidae, ping

pong, sort of. i'm eating dinner & watching simpsons in a minute, but i'll check in every now & then

CatCat: You said http://faux.uwcs.co.uk/dupscan2.html was ugly (I think), what'd you prefer?

looked into that rip issue a bit more last night

i ripped that last track with plextools pro,a nd it agreed with EAC

interesting part, though is DMC produced the same output as cdparanoia

and it said it was accurate according to accurip with confidence 2

which is what EAC reported for the output it produced

so i am confused as to what is going on

FauxFaux: http://chatlogs.musicbrainz.org/2007/2007-09/20...

uh, that's weord

weird

(you SO missread that)

i guess i need to dig into exactly what those bytes near the end of the file are

yeah

ugh ENOUGH with the ean/paranoia7whatcamafuripping

i've done some testing today myself with ripping badly damaged cds

getting weird results myself

never tried much with damaged cds

i'm completely anal about storing my cds :/

i keep them in mylar sleeves

burned a cd, dragged it around on the floor, dusted it off and put it in the drive

FauxFaux: what i meant was that the report iself is pretty slick and i dig it

the amounts of dupliated data made me cry

i found a plexwriter pro on clearance somewhere and picked it up very cheaply

ripped with cdex, got some errors and a really nasty audible error in the middle of the song, but except for that it was ~flawless

something tells me that if software is *hurting* hardware you shouldnt be usign said software

catcat, i meant i'd be afraid dirt or similar were on the cd

and would muck up the drive

also what are you guys *doing* with your discs? using them as costers for dinks or what?

ripped it with eac and the really nasty audible error wasn't there, and the rest was just like cdex

CatCat: Aha, okay. :) The page isn't particularily pretty anyway, but I couldn't find anything that hurt my eyes less, I was wondering if you had suggestions.

so i ripped it again with cdex and the nasty audible error was gone, producing same output as eac

i'd like to dig into this more when i have some free time

i really would like to produce something that is as good as EAC that doesn't need windows

fauxy: useful woudø be a 'merge' link

found some pretty good info

but that woudl have to have args for all the dupes aid

wich in the xase of 4 discs wcoudl get.. ug hard?

CatCat: (Afaik) there's no way to link directly to a merge-both page. I'll have a look. :/

cdparanoia is pretty close, but it doesn't do gap detection yet, and it can't figure out the correct read offset correction

cdrdao will do gap detection already, so code can be reused from there

and there's code floating around to pull read offset from the accuraterip db

ojnkpjg: well, i damaged the cd even further, ripping now takes 3-4 hours

and code to calculate the accuraterip checksum for comparison

hehe

first round cdparanoia won, but i cranked up the settings on eac a bit to give it another go

i'm more interested in the case where the cd is in pretty much mint condition

i'd like to get a libparanoia based app to the point where it agrees with plextools pro (and eac)

i'd trust the former more

but i've never seen differences with undamaged discs

ojnkpjg: did you use the -O switch with cdparanoia (sample offset)?

yep

btw Rubyripper is being developped as a replacement to EAC under unix

i've heard of it, but i've never checked it out

and i don't know ruby

but i guess it's probably learnable in an afternooon

thanks

oh yeah it's simple enough

it's coded *in* ruby but you don't have to know Ruby obviously

well, if i wanted to work on it, i would :)

right

its algorithm has been described extensively on hydrogenaudio.org if you want to do something similar

i've been reading through the ha forums a bit

no point duplicating effort