Whenever we get introduced to a site that features some kind of “digital fingerprinting technology” to power a music recommendation engine, it’s hard not to be skeptical. Fingerprinting works well for identifying the same song in multiple places – for example, duplicate songs in a music library. But when it comes to music recommendations, automated systems rarely work well, which is why Pandora relies on a team of 50+ analysts to drive its recommendation engine, and most other sites rely on social suggestions and meta data.
Today sees the private beta launch of Mufin, an automated music recommendation engine that actually seems to work(and it’s really, really cool when it does). The technology is sophisticated, but this isn’t surprisingly giving Mufin’s pedigree: the site is an offshoot from the Fraunhofer Institute, the German research organization that originally created the ubiquitous MP3 audio compression algorithm. You can grab an invite to the private beta here.
After searching the site’s 3.5 million song database for one of their favorite songs, users can press “Similar Tracks” to view a short list of songs that have been determined by the site’s powerful algorithm to contain similar characteristics. The algorithm analyzes 40 characteristics of each song, including tempo, sound density, and variety of other factors. Users can search for songs both from Mufin’s homepage and through an embeddable MySpace widget (with Facebook on the way).
In practice, these recommendations seem to work very well – at least more than 75% of the time in my testing. Many of my searches for popular songs (such as Here Comes The Sun) yielded results that sounded very similar, though many of these matches were in a different language or were recorded by obscure bands. To rule out the possibility of pre-determined matches, I tried a few more obscure recordings. The Darkness’ timeless classic “I Believe In A Thing Called Love” yielded “Sucker” by Europe. And Kyu Sakamoto’s “Sukiyaki” was matched with “Twilight Time” from The Platters – a song I’ve never heard of, but is unquestionably similar. Note: I’ve linked to YouTube clips for the above songs so you can hear the similarities – there doesn’t seem to be a way to embed Mufin results yet.
Unfortunately, while the technology behind the site is impressive, it still needs quite a bit of work. Many of the matches seem to be accurate, but the site offers no explanation as to why a given song was recommended. Mufin says that it will be including Pandora-like explanations detailing which characteristics have been deemed similar between each song, but for now you’re left in the dark.
The site also has a number of licensing issues to work out. Mufin has partnered with Universal Music Group, but many songs in the database are lacking preview clips, which make some matches useless as there’s no easy way to listen to them. Other songs in the catalog have previews but are missing links to online music stores like iTunes or Amazon, so there’s no convenient way to buy them legally. And results are riddled with duplicates from ‘Greatest Hits’ albums.
Beyond these issues, Mufin has a lot of potential, and could provide a viable alternative to Pandora or sites like Last.fm once it cleans up its interface and squares away its licensing issues.
Got any invitation codes?
Yep, hit the link in the post.
Doh. Thanks, sorry.
Can I have a muffin please?
If this company came into my office for an interview i would not hire and recommend that they go home and change the clothes (domain) and Logo there wearing. This domain and logo hurts.
DomainerLocator.com
god what is it with this locator dude? look mate, your clothery is way too long recommmmmennddationnnnLocator.com and ur cussing off mufin.com? I’ve been to ur site a few times and its just kak, like, what the hell is going on there, what the fuck is it about? And what about ur logo? its clipart and not even good, its like some web 1.0 pic
How much have u spent on all your locator domains, maybe 5K, and how much more until u buy the whole dictionary? another 50K. why dont u use that money to buy loc8tor.com or something else unique and use the sub-domains. The point is, if ur wierd concept (and ive your videos, your definately off ur rocker) does become big, then locator.com will simply kane u off, they got the domain and is the place people will think is the starting point.
You’ve got a lot of energy, but it is so misguided.
(and no, i don’t want to go and duel it out with u on seesmic, i’ve got better things to do)
haye, I sent a request for the invitation code few days ago but I have not got that yet,how long do they usually take moreover don’t you think 3.8 M songs are not good enough ?
Keep in mind that they’re just getting started. This is a private beta site after all. And 3.8 million songs sounds like a lot to me.
iam thinking of a itunes plugin that will automatially recommend similar music. can the same idea is applicable to videos?
Can I have a mufin please?
Looks nteresting. Anyone know when Instinctiv plans to launch their shuffle on the new iPhones? Really miss that app.
Can I have muffin please?
Maybe its just me (my comp) but the playback quality is terrible on the recommended songs. Not impressed by the service over all. They’ve got a lot of work to do if they’re looking to compete with pandora and last.fm
Hardly accurate. I’m not sure what set of songs the reviewer used to test, but I picked Oasis and the Verve (two bands that would be considered ‘Brit Rock’ or ‘Britpop’) and most of my matches were heavy metal. One of my matches, matched at 91%, was a death metal song.
sigh… gave it a whirl. used the invite codes.. lets be honest folks:
FAIL
Are these guys eastern bloc? Love the grammar of their reply:
Hi,
really cool that you’re interested in our private beta. We’re working some things and will sent you your personal invite code for our music discovery engine soon. In the meantime, you can keep updated on the things happening at mufin on our blog: at http://blog.mufin.com
Your mufin team
They’re from Germany.
How good is your German, RL?
(For the 2nd time, as my previous post vanished into the ether)
I just tested this briefly with 3 artists and found it not to work very well at all.
First off, my current new favourite band, The Faint.
http://beta.muf...872/?src=search
No recommendations. Both Reverend and the Makers and The Presets sound similar to The Faint (in my opinion), that’s what I would personally recommend as similar music. Mufin seems have have both these artists, so why aren’t they recommended.
Next, tried 505 by the Arctic Monkeys. Such a great track.
http://beta.muf...kdfsf/23108794/
No audio for first three recommendations, so can’t tell. But next three are all totally different genres of music. Maybe to an algorithm they sound the same, or technically speaking they are similar, but as a fan of music – um, nope, no thanks.
Finally, my FAVOURITE!
Kings of Leon. Searched for Crawl off their new (fantastic!) album. And they recommended other tracks for me that are similar. Because. They are FROM THE SAME ARTIST! Genius.
So, let’s try an older Kings of Leon album.
http://beta.muf...058/?src=search
You ready. A rubbish reggae song (089). And, the clincher. A 100% crap Austrian euro-dance track. I mean, come on. Seriously?
Sorry, forgot to mention, the Austrian techno song is about 5 down, it’s called Einmal Entschieden. It’s amazing. Truly.
Entered three different songs and none were in the database, so no matches were found. I went with Wilco for one song. Can’t be that obscure…
Hmm, which song did you try? Wilco worked for me.
I love these types of sites and can’t see why there are such dumb legal issues associated with them….I’ve actually gone out and bought CD’s that Last.fm turned me on to, and have been exposed to a lot of new and excellent music, something I’m sure musician’s would welcome for exposue….
How does this compare to MusicIP in terms of recommendations (http://www.musicip.com)
Try out our newer site, Bob: http://mydj.musicip.com
It’s been up in beta for a month or so, and we’re working to make it better all the time, but it lets you discover music and also create acoustics-based playlists. And, we go ahead and give you access to tweak the rules that are used in driving the discovery and playlisting engines via the ‘DJs’.
Let me know what you think!
Oh… we also have our technology rolling out on a few new phones over the next month or two. Keep an eye out for MyDJ!
Ok, Mufin seems to think Modest Mouse sounds like Kenny Chesney. Does that seem right to you?
Which song were you trying? Sometimes the songs are definitely a different style, but might have a similar beat or instruments.
Satin in a Coffin. And the Kenny Chesney song it matched it with sounds nothing like it. I don’t remember what Chesney song it was. I do actually like Kenny Chesney, though, so maybe it was reading my thoughts.
Tried five artists, then five songs, not impressed. The Decemberists came back with Van Morrison. I can stretch my mind to come up with a connection if I had to (I don’t want to). But the point is, this is a music DISCOVERY tool. Van Morrison? Please.
Does anyone else think that logo reads more like Mifir than Mufin….?
Jason…Great post! Could you hit me up via email, I lost you on my contact list
Music is going places on the Web obviously and you are quite tuned in I see as well.
Keep doing these great posts Jason and let me intro you to some people who will amaze you too.
Always,
Phil
Sorry, but David Hasselhoff’s “Flying On The Wings of Tenderness” is not an 80% match to the White Stripes’ “Fell In Love With A Girl”.
I actually found some decent matches for The Hoff’s rendition of “Raindrops Keep Falling On My Head”.
Tried it and I just dont this kind of system works based on the reasons I like music. When I say I like a song like “Stan” by Eminem, its because of how well the story was told, the unique structure to the song as a series of letters. Do these programs actually study these kinds of things? I doubt it. Although systems like Pandora and Hit Song Science seem to work well in some instances, overall there is still much more work to do to make these algorithms stick. The best approach to me is one like iLike, where you use people’s feedback as the main engine for recommendation, since this helps with the intangibles of why people like certain songs.
Thought this was horrible too – tried some instrumental music (techno, improvised) and got back things like Georgie Fame and The Corrs. Pretty terrible if your game is recommendations, although I do quite like the cleanliness and relative simplicity of it all, which may be worth something given how nasty the new last.fm is.
I don’t think so they are better than last.fm .
They are different.
I ♥ Last.fm but it sucks as a music discovery tool (is more for social networking). Pandora is way better, but I can’t use it in my country
My experience with this has not been good:
http://www.9gia...teps.com/?p=963
George
I gave it a go over my lunch break. Punched in Nine Inch Nails, scrolled through some of the songs until I found something that I figured would be interesting to get results for.
For Nine Inch Nail’s “Closer” (Live from And All That Could Have Been), it’s top recommendation was Britney Spears’ “Oops… I Did It Again”.
Sounds like a great way to find some great hidden bands that I can get friends into. Stiff competition though, not sure how they are going to make a name for themselves.
Craig
http://www.budgetpulse.com
doesn’t look too appealing..
http://gatesand...s.blogspot.com/
Look forward to trying it out. Always down for a new way to discover new songs.
I’ve found that its often best to compare songs rather than bands because if you consider a band’s entire collection, styles are often very unique, and much harder to associate with each other.
I really wanted this to be good: for one thing a proper music recommendation system would be the final nail in the RIAA’s coffin and put an end to their crap about how the net isn’t as good at distributing music as they are. But Muffin, very sadly, it shit.
I don’t know what last.fm waas like when it launched but this is very underwhelming. Data coming from people is more useful at this stage I think. I don’t want music that is the same, I want music that appeals to my tastes – this is not necessarily the same thing.
nice one
mufi is like cuffin
Thanks for the invite. Just gave it a try.
Artist: Oasis
Song: Wonderwall
Results: “Sorry, no tracks similar to Wonderwall available”
Verdict: Fail
Artist: Green Day
Song: Minority
Results: First three are some German industrial techno bands I’ve never heard of. Then…
#4. “Our Time Now” by Plain White T’s
#5. More German stuff
#6. “Crossroads” by Hazen Street
#7. “Seasons” by Good Charlotte
Verdict: Weak
Then… over at ZuKool:
Artist: Green Day
Song: Minority
Results:
#1. The Clash “The Guns of Brixton
#2. Crass “Where Next Columbus”
#3. The Ramones “Needles Pins”
#4. Crass “Poison in a Pretty Pill”
#5. Crass “Systematic Death”
#6. The Ramones “I Can`t Make It On Time”
#7. The Buzzcocks “What Do I Get”
#8. The Buzzcocks “Nostalgia”
#9. Green Day “All the Time”
#10. The Buzzcocks “You Know You Can`t Help It”
Verdict: Winner
ZuKool seems to be the better beta brutha.
(And nope… I’m affiliated to ZuKool in anyway.)
uhh… that was meant to read “I’m not affiliated to ZuKool…”
so far this blows… searched for lots of well known alternative rock songs and I got a lot of blank results and a few results only returning the song I searched on. lame! (props to Kim Thayil)
Yeah, I had a pretty bad experience with Mufin as well (disclosure: I am co-founder of the MixZing music discovery solution). I just have a fundamental problem with these technical affinity approaches. As others have said, how can you really capture context and mood by looking at melodies, beats, etc.?
We take the collaborative filtering approach at MixZing, using tons of real user playlists as input, and the recommendations we give are specific to a given playlist, since otherwise they’d have no context. Check out our iTunes plug-in if you’re interested (sorry, that’s my shameless plug-in plug).
I tried Jonathan Coulton’s Code Monkey: no matches. Then Owner of a Lonely Heart: first three matches were the same song off other compilation albums. Finally, clicked the link for NIN’s Closer and saw the top match: Britney Spears’ Oops.
I haven’t been this disappointed in a new product since Cuil launched.
Complete fail. I’m not so much disappointed in this Mufin service as I am in TechCrunch’s review. I always thought this blog told it straight. Did they actually use this service? How can everyone else here have a 90% fail rate on matching songs and TC get such good results?
its kind of obvious, the similarity is based on measurable acoustic criteria of course it does not recognize a musical genre, metatags or social classifications. sure, its kind of dumb to build a whole service around merely technical data without combining it with other strategies.
For a simple music search engine where you can listen and download directly from one site without having to register try http://www.plorf.com
But plorf.com doesn’t find matching music! Mufin is about finding matching music – on the net or offline with the player on the hard disc!