Last.fm Plans Custom Music Video Channels
Corey writes "CNet's Crave is reporting that the popular Web 2.0 music site Last.fm is planning to launch a video-on-demand service that dynamically creates a custom video channel for users in the same way it currently does with music. Read/Write Web also cites a recent press release quoting directors at Last.fm as saying they plan to host every music video ever created. This could well turn out to be the MTV of Web 2.0."
I still get server overload errors on Last.fm (just had one, in fact), and that's with just audio (plus track listing data, of course). They'll need a big investment in infrastructure to serve up video reliably.
Quidnam Latine loqui modo coepi?
Last.fm is UK based. So presumably has to deal with the BPI's abominable licencing.
=DIVIDE BY CUCUMBER ERROR: REINSTALL UNIVERSE AND REBOOT=
I was just looking for an internet radio site... I'm working from home today and I couldn't find the power cord to the old portable radio. last.fm seems cool -- based on my first try. When you first login, it asks you for what band you're interested in listening to. I figured I'd give it a challenge, so I tried Mr. Bungle. A decent song came right up (Backstrokin) and it also gave me some decent suggestions for related bands -- some I hadn't heard of. The next song is right on key, so I'll keep listening for a while.
They have a horrible system for recommending similar music to users. They need to adopt the Music Genome Project from Pandora or something similar that they can create with their large and pretty active user base.
I use last.fm all the time and I love it, but not for what it is billed for. I am able to track my listening, which is cool because, dude numbers.... right? Also they have information on artists a click away from my page along with a calendar for events coming to my area. They use their database well in many ways, but they have yet to come up with a system for recommending new music that is good. Currently, it boils down to, if you like Artist A and these other 1000 people like Artist A, then you must like some of the other stuff they like. Then the bring in tags and produce some list that is bogus. Given they make money from selling CD's and getting a cut, you would think a large focus would be put on a system where people would actually find stuff they like when searching for similar sounding stuff.
As for the music videos - I do not care to much for the idea, but I see the draw for others. They have improved their site a great deal over the last few years and it shows no sign of slowing down, which is good. I just wish they would revisit the core elements of the site for a change and give us a system that works... it is, after all, what the site was meant to be about in the first place.
Invexi - a Phoenix, AZ based web design and web development company.
They have an excellent database, very extensive, and host lots of videos (certainly I haven't been able to think of one that don't have yet)
I think they are major label, too, so probably have worked out whatever licensing kinks there may be.
Point is, this is not a newsworthy item - lots of similar sites exist and have existed for a while.
Contemplate the marvel that is existence, and rejoice that you are able to do so.
Comment removed based on user account deletion
As much as I appreciate Pandora's ambitions, their recommendation system is far from ideal. I can only tolerate Pandora radio when I'm in a real funk where I don't want to listen to anything in my collection, and even then, it never takes long for the database to actually offend me with its poor recommendations. The flaws come up more in some genres than others; one of my biggest gripes is that a lot of the harder music doesn't seem to be labelled and categorized by anyone who actually is a fan of hard rock. Rather, it seems like after a certain threshold, whoever's categorizing it throws their hands up and says,"It's all just noise to me!" and just throws the crap and the good together willy nilly.
Also, the "Add Music to This Station" functionality is really poorly implemented. It seems to treat each seed seperately and just rotate through which seed's recommendations it's pulling recommendations from, rather than doing anything sensible like finding common themes and traits among your favorite artists all together and then basing recommendations off the pooled knowledge. I can understand some people liking the current system (particularly fans who stick to one genre), but I'd been hoping for a way to tell the system,"All your recommendations for [band] suck, pay more attention to the aspects [band] shares with [other band]".
...which isn't to say that last.fm's popularity-based system doesn't also hold annoyances. If you like a single band that's part of a larger scene, but don't like the rest of the scene, you just get flooded with crap from that scene no matter where you turn (for values of "scene" spanning everything from cliquish sub-genre to country of origin to decade of origin).
Basically all automated recommendation systems suck at this point and I'll never find good new music ever. DX
Maybe you just don't get the core elements of the site, but it has always been a social network site based on similar music tastes. The Pandora project is an interesting idea in that it algorithmically determines music similarity (not hand coded like a poster above mentioned(?)) based on attributes such as tempo, etc. Last.fm, on the other hand, rates similarity based on what people are listening to. Meaning, Kelly Clarkson and The Killers (for example) do not have any musical attributes in common, but sure as shit the same people that listen to one, listen to the other. It breaks down with popular media because lots of different listener types will listen to the new rap/dance song, but again as someone said above, it works for things not so much in the popular culture (Ska, Screamo, Reggae, House, Funk, etc.).
The social network has been, and always will be, thier core element. Thier recommendation system is as good as the people listening to the music.