PDA

View Full Version : The CDDB are the new FBI.



Woebot
19-10-2004, 10:02 AM
This weekend I hand-recorded a vinyl LP onto my CD recorder (the better to be able to share it with my chums). Obv the whole thing was engulfed in an authentic low crackle and on two occasions the front-end of the track was smeared by the sound of my technics deck accelerating into it. It's a really fiddly proceedure. The recording is very good if a little flawed, no matter.

Monday morning in the office i slip the CD into the iTunes at work and, to my sheer disbelief, find that the CDDB (or Gracenote or whatever the F theyre calling themselves today) had correctly identified my hamfisted recording. All tracks meticulously tagged in Portuguese! How the hell did that happen? I thought these things were roughly ID'd by the track lengths!?! Are they analysing the waveforms? Spooky!

MolexRoots
19-10-2004, 12:52 PM
That is very scary O_o

I'm going to try that later today.

Woebot
19-10-2004, 01:06 PM
...and it just happened with the other CD I recorded as well (Jorge Ben's "Tabula Esmerelda") Terrible crackle etc. They're in league with the devil.

gragy10
19-10-2004, 04:04 PM
CDDB is terrifying - I've had it tag bootlegs and 12" rips before. Guess it's a very very big database

stelfox
19-10-2004, 04:18 PM
it's better when it gets things wildly wrong. i have had microhouse records identified as garth brooks tunes before now. wish that could explain away the marilyn manson listings on my hard drive, but sadly it doesn't ...

Jamie S
20-10-2004, 12:59 PM
Maybe it's something like this

http://www.shazam.com/uk/do/home

which is truly amazing. We tried it out while listening to an electro-clash mixtape that someone had made me and it was identifying stuff just from the thumping beats before any melodic action had started.

Someone was telling me it started out as an academic project, but it doesn't seem to have anything on the history of it on the site. I don't understand the economics of it at all. For it to be any good it has to be pretty comprehensive so you've got your entire costs frontloaded before you can make a penny.

My mate works at the Sound Archive of the British Library, where they're debating how to go with recording stuff onto digital media. If that all got loaded, catalogued, searchable... frightening really.

MolexRoots
20-10-2004, 01:03 PM
I've always been skeptical about Shazam. Never used it, btw. Don't mobile phone microphones have a Very limited frequency range? I forget the exact number but I know it's smaller than a normal house phone. Trying to play music through that is just a shrill noise, especially in a club where music is so loud that it distorts on a weak receiver like a mobile phone.

Does not compute. Bzzt.

Jamie S
20-10-2004, 01:12 PM
I was totally skeptical too (and drunk). It seems astonishing that something so crappy as a mobile could pick up a decent enough recording to match a waveform or whatever. But really, it works!

Boomp boomp boomp boomp. 2 minutes later you get a text saying FC Kahuna 'Machine Says Yes'

If anyone does know how they do it, I would be fascinated to find out

grimly fiendish
21-10-2004, 10:41 AM
yes! i was thinking about that - shazam - the other morning, basically because i had need of it but couldn't remember for the life of me what it was called or what the number was. and i figured that, because i hadn't seen any promotion for it for ages, it had gone under. obviously not.

how the *hell* does it work? how do they maintain the database? how many hits is it getting? someone must know. i'm fascinated.

hint
21-10-2004, 12:47 PM
I don't think it's anything to be sceptical about - if they don't tag the track you're after, you don't pay, if I recall correctly.... but it certainly works very well for stuff they have in their database.


Shazam's service is built around a proprietary pattern recognition technology (patent-pending) that can identify recorded audio even under noisy conditions.

I guess the fact that the technology is patent-pending means that they are deliberately guarded about letting any of the details of their systems leak out into the public domain. All I know is that they tour round various music libraries (publishers / labels / whatever) and run through their entire catalogues (CD only, last time I checked). They literally stick a CD in their machine, read and process this mysterious "pattern" data, and store it in their datatbase.

They recently reported that they'd successfully tagged 4 million tracks in the UK, so it's clearly a success - a great idea properly executed, even though it seems like black magic, considering the various issues of sound quality, pitch variation etc you get in in a club environment! We're talking once-in-a-lifetime stroke of genius type business here. Good luck to 'em.