8

I posted a video on YouTube (SFW) nearly two years ago, with an ambient soundtrack taken from an Aphex Twin song (which I credited at the end of the video). After nearly 3000 views, it has had the audio stripped out and a copyright notice attached on behalf of the content owner (WMG). It says, "Matched third-party content." That's fair enough.

What fascinates me is how did they work out that the audio was, well, dodgy? Does anyone know how that happens? It can't be a manual process, can it? Someone playing all the videos and suddenly going, "Hey! I know that song!"?

EDIT: See 'Google seeks to turn a profit from YouTube copyright clashes' which seems to support Olly's answer, which is why I'm accepting that one.

migrated from superuser.com Nov 27 '11 at 8:34

This question came from our site for computer enthusiasts and power users.

5

My guess is that they're comparing the audio in the videos to some sort of acoustic fingerprint database. When they find a match to something they know is copyright, they disable the audio.

  • Shazam, mentioned already, uses acoustic fingerprinting, so I guess this echoes izb's answer – Ralph Lavelle Oct 8 '09 at 1:16
  • Yes, I guess so, except izb seems bewildered by how it works :) – Olly Hodgson Oct 8 '09 at 11:29
4

It could have been someone (or more likely several people) reporting the video - you've had 3000 views, which is enough for someone to recognise the soundtrack and realise that you didn't have a licence.

4

Could be something like Shazam, which always bewilders me in how it works. You phone a number, let the service listen to the music you're hearing with your phone and in a minute or so you get a text message with the artist and track. It's frankly quite spooky. I guess Google could be doing something similar, or using the technology.

4

Youtube does a lot of things to analyse the audio. One is speech to text to display the right ads. Another happens to be matching some sort of hash to identify music that's owned by the labels.

4

YouTube uses a system called ContentID. Basically the moment the file is uploaded, it's analyzed and matched against a database of known material submitted by the record labels, along with a policy of what needs to be done when a match is found.

Here's a TED video which explains the entire thing very well.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.