Hi Lukas,
Do you have some (good) news about to use Chromaprint with AcoustID (Business version) for “Real-Time stream identification” ?
Thanks for your help
Best Regards,
Chris
It’s a topic for From what i’ve heard from previous discussions it is not possible to do with a stream.
acoustid is designed to process static files with a known starting point.
With radio streams you don’t know where the track begins so if you calculate the acoustid with a different stating a fraction of a second later you will get a different id.
2 Likes
Yeah I’d be surprised to see AcoustID get into stream identification. Though, who knows, it’s a field full of proprietary implementations, maybe it’d be somehow beneficial if it did?
The trouble is that it’d be a completely different technology. Almost any scheme would require voluntary participation of streamers. It’d either be audio watermarking [which uses a repeating, difficult-or-impossible-to-hear audio signal, unlike Chromaprint which tries to fingerprint what can be heard] or a dedicated metadata tag. Again, many proprietary implementations of these already exist.
(And, quite often in these implementations, their main intention is to enable DRM and third-party tracking.)
What may be useful is acousticbrainz.org and trying to compare based on some of the calculations that it does.
Potentially you could combine enough of the results of these algorithms to find your recording.
This is a collection of algorithms that you can run over a recordng and has things such as bpm, key, mood, dance ability and all sorts of things.
Some of these should be within a similar range throughout the recording and if you have enough parameters you may be able to compare enough to get a reasonable guess.