Need to find the timestamp of an audio clip inside the audio file from which it originated
February 6, 2012 3:26 PM   Subscribe

Does software exist that can search for an audio clip inside the larger audio file and return the timestamp?

Is there an application or a website where I can feed in two audio files--one being a clip from the other--and it will return the timestamp at which the smaller occurs inside the larger?
posted by raddevon to Computers & Internet (5 answers total)
I've found a company that does audio search within a library, but not inside individual clips or streams.

As I think about it, it's not a straight-forward problem to solve.

Can you tell us about the audio clip you're using as a search parameter? Is it, like, a separate recording? Or, was it ripped from the digital audio stream? Or is it a bit-perfect copy of the PCM as decoded from the file you're searching in?

If the search clip's original source was the (unmodified) audio file in question, then you might be able to get away with some sort of simple program to decode the audio and do time sequence or string search. You could probably find somebody to hack that up for a grand or two.

But, if it's a separate original recording, or the search clip has been altered from the original stream, you're going to need something sophisticated. And none of my googling is finding such a thing.
posted by Netzapper at 3:48 PM on February 6, 2012

I wonder if shazam has this ability?
posted by roboton666 at 6:24 PM on February 6, 2012

I don' think Shazam will work unless the clip that raddevon is trying to find is in a song. (Just in case here's a list of services that are similar to Shazam.)

I don't think that you're going to find this ready to go. I do have two ideas for you though:

If your audio clips are of speech you might be able to get away with using some kind of open source speech to text on both of your samples and then matching up based on the recognized texts.

If speed isn't a requirement and you're willing to throw a little bit of money at this you could probably use Mechanical Turk.
posted by gregr at 6:39 PM on February 6, 2012

The clip was cut from the larger file. They are MP3s so I'm sure the clip is slightly more compressed than the original.

The clips are of speech. I might try the speech-to-text option. Mechanical Turk will not work for me in this case as the material is confidential. Great suggestion, though!
posted by raddevon at 6:58 PM on February 6, 2012

Have you tried adding the clip as a movie to youtube then running the autocaps procedure? That might give you the timecode of when the speech you're looking for starts.

This article is from 2009, but I'm assuming it's still available.
posted by urbanwhaleshark at 5:28 PM on February 7, 2012

« Older Why can't I be Sherlock, Indiana and Marie Curie...   |   Hulk Smash International Viewer Newer »
This thread is closed to new comments.