Things are going fairly well for a first release of this next week. Here's a WIP screenshot... http://www.flickr.com/photos/dan-gravell/6761376997/

Any comments welcome, although I know there's a lot one simple screenie doesn't say ;-)

@jake see also this idea: http://ideas.blisshq.com/forums/21939-bliss/suggestions/2458603-use-the-folder-and-file-name-to-help-identify-song however I'd rather work on the acoustic fingerprinting approach first.

This would be useful. At the moment untagged files are ignored. The file names could be used to identify the music?

Acoustic fingerprinting is now in for looking up missing information (cover art, year, genre) - http://www.blisshq.com/music-library-management-blog/2011/12/14/new-release-20111206-padding-file-org-acoustic-fingerprints/index.html . The next step is to provide a way to identify untagged files and fingerprint them, offering them as fixes in the UI.

Currently I'm working on integrating acoustic fingerprints with existing functions of cover art, genre and year lookup for the next beta. I think this should provide higher accuracy and slightly better coverage. This should be released in the middle of December.

Once that's done and stable I'll look at the tagging for untagged files.

Sounds great, presently I always use MusicBrainz Picard to initially tag new music, then I get Bliss to scan and correct any compliance issues and also add in the coverart which results in a very accurate music library.
The other big advantage with this is Picard also adds the MusicBrainz identification tags to every song, which enables easier identification of the files next time they are scanned.
On a side note I'm also using the lastfmplus plugin for Picard (http://musicbrainz.org/doc/MusicBrainz_Picard/Documentation/Plugins/Lastfmplus) to populate multiple genres into the single field, incorporating data from LastFM and separating it with a delimiter. Then because I use Logitech Media Server for searching and playback of music this allows me to classify songs into the multiple genres spliting them out using the delimiter.

I've agreed with Acoustid the use of their service. Initially I will integrate it for improving cover art lookups. I can then expand that to looking up all tags (genre, year etc). Then, the piece of work this represents will be the tagging of untagged music.

I hope this represents the intent of this idea correctly...

An update on this... last week I ran a feasibility study using Acoustid.

The results were encouraging. Of a sample collection of just over 5,000 tracks it correctly identified 89% of them. Accuracy is superb. I think this is worth taking further.

There are two main uses I can think of right now. First, it's a solution to a problem in its own right: auto-tagging untagged music. Secondly (and this isn't so much what this idea is about, but it's worth noting anyway), it's a more accurate way of looking up information for music that already has tags.

Another possibility, which may or may not be related to Acoustid - http://echoprint.me/

I just wanted to update everyone on this. Implementing this used to require a contract with a company to pass the fingerprint for identification, and then use the ID to refer to another music data source like MusicBrainz. I have contacted some companies about this, but they were inflexible to the contract I wanted to setup.

Recently I discovered Acoustid which is an open database of audio fingerprints - see here: http://acoustid.org/

I haven't done any POC yet but I'm hoping it might work out...

@justin - no, but it certainly got the blood pumping.

can you please identify the song in this video?

http://www.youtube.com/watch?v=YbCMb1kB-tw&feature=related

Fingerprinting would be great

Along with Bliss, I've used Picard, Magic MP3, MP3Tag DBPoweramp and TuneUP(I tunes add on).
I've found that Bliss is dead on with what it's set to do.

Fingerprints are a really neat way to go.
They're only as good as the database holding them.
I guess musicbrainz is the major database. I think they
need to expand it. ( Sounds like something the Library of Congress should maintain )

And then, as you had mentioned if you're going to automatically fill in the
metadata, it needs to be correct.

There's nothing like being a DJ and in a pinch trying to grab for a
Ray Charles track .... and you wind up with Willie Nelson ( full LP mistagged ).

:) Cheers.

Sorry for late reply, my server had to be reinstalled.
Yes, audio fingerprinting would be a good feature. It could verify already tagged files as well as files without tags. Great for QA purposes to clean up tagging on large mp3 collections.

Ok, so you would want tracks and albums identified when they don't have any tags at all? That's something I've wanted to do for a while.

I imagine Magic MP3 uses audio fingerprinting works, a little like Picard the MusicBrainz tagger.