Got some mp3 with completely no info inside them. filename is a random string of characters, cause i exported them from ifunbox (an iphone utility) when i made a backup of my iphone before i updated to a newer ios version way back. Only usable info is in track name where it only displays Track number. So for example i have a folder of 55 orphaned files with names like “[untitled][2164][i-funbox.com].mp3” and tag in title field is “Track 6”. Same structure goes for the rest. And theyre not all from one single album, i got like 3-4 mp3 for its track number, eg 4 random named mp3 with Track 1 in title tag.
Run a scan and musicbrainz only recognises 16 of them. Even after i input a fingerprint generated by my musicbrainz login.
Shazam easily recognised one of the unamed tracks, but i cant shazam the whole folder, and even so basically i’m looking for a futureproof way cause even after i take care of these file i will have to clean my whole music library and hundreds more like these will popup eventually
You probably don’t want to hear this: It’s almost impossible.
Without metadata and without references to file names, no software is able to guess your songs with high accuracy. With the help of Acoustid/Fingerprints there will be some random hits, especially with very popular songs.
But all songs that you don’t recognize yourself while playing them will be extremely difficult to guess.
I figured as such, but how does shazam do this. Cant this be incorporated into software that does this with stored files. I cant keep playing files into shazam to handle a bulk of unknown media
If you are interested in the details, you can read an academic paper published by Avery Li-Chun Wang to describe how Shazam works.
A shortest possible description:
Shazam creates a spectrogram for each song in its database - a graph that plots three dimensions of music: frequency vs. amplitude vs. time.
Well sure it does this somehow. But what my question implied was how is it able to do it, and MusicBrain or other various apps dont include such code, where you click to submit and get similar results
I think Shazam (through Gracenote) has access to a database.
And that access is heavily restricted in efforts to maintain the scarcity and financial value of that database.
Pretty much. AcoustID does what Shazam does on full files though, it’s just (same as with Shazam) you depend on the stuff you have being already in the actual database. Shazam has a much stronger financial incentive to have all popular music in their database, since that’s their main source of income. So while it often fails with obscure music, it almost always works with more popular stuff. AcoustID mostly depends on us music nerds, so I suspect it has more hits on absurdly obscure stuff, but might be missing a fair amount of popular music no user happens to actually listen to.
In a case like yours try lowering the track matching threshold in Options > Advanced > Matching.
If you do AcoustId lookups that acts as kind of a sanity check to avoid total mismatches by comparing existing metadata against the found data and calculating a similarly. If the similarity is lower than the threshold no match is done.
The default is already rather low at 40%, but as you have really bad existing try lowering it to maybe even 0.
Also some time ago sometimes the AcoustId servers had some trouble. In such a case just retrying scan on the already scanned files might give additional results.
Hi. I have a similar situation to the above. I have a load of old audio CDs containing tracks that I ripped from vinyl singles 10+ years ago. I’d now like to rip the CDs to MP3/FLAC without the hassle of re-recording from the original vinyl. The main difference is that I DO know what the tracks are, but it would be great if I could select the metadata from suggestions rather than having to type each track in, one by one.
The tracks are generally well-known, and Shazam / Soundhound identify them within seconds. But I’m not getting a single hit when selecting ‘Scan’ in Picard, even with the thresholds set to zero - not even incorrect hits or suggestions. Nothing. I would have expected some results, even if they were wrong.
If I use the Lookup feature, I’m just getting lists of tracks that are all called “Track 01” rather than matching the length of the track.
So I’m not sure if what I am trying to do is impossible, whether I’m doing it wrong, or whether the servers are having some problems - could anyone provide any suggestions please? Thanks!
What are the files currently tagged as? Try using something like MP3TAG to manually set the track names. That would then allow the < Lookup > button to work better. Lookup just reads tags.
AcoustID scans should be attempting to match your audio, but if you happen to have something less popular, or your recordings start at a slightly different point, then this may be why the < SCAN > button is failing to find matches.
You may have more luck working the other way round. Go find the SINGLEs in the database. Use the < Lookup in Browser > button, then navigate to your actual Single. There is a GREEN < TAGGER > button that pushes that Single back to Picard. Now you can drag and drop your tracks from left to right and match the single.
