cheesebikini?: The iPod Killer?

October 26, 2002

The iPod Killer?

ipod killer Imagine listening to the mp3 music files stored on your PC or server, from anywhere, via your mobile phone/device. Just browse through your music via your device's display, choose mp3's to stream to the device, plug in your headphones and groove.

Why lug around gigs of expensive portable storage space when the music lives at your house, or on rented Web server space? Why hassle with multiple copies of your music, one on each device and computer that you use, when that music can live in one place?

Brad Lauster's recent blog entry got me fired up about this. Someday people will store their music and other files centrally and use them via multiple devices in multiple places; that's almost a given. But don't we have what it takes to make that someday today? Don't plenty of people who regularly listen to mp3s have devices that can support such a system?

Someone just needs to build device-based client software that allows folks to choose and stream down their songs, and the corresponding server software that talks to the device and serves up the chosen songs from the PC back home.

So what do you think -- is the bandwidth cheap enough to make this sort of streaming practical yet? Certainly during weekends, when minutes are free, no? Shoutcast streams mp3s effectively from PC to PC, even over modem connections, so I think we already have sufficient bandwidth. The device-side UI would be quite a challenge, but what a great capability this would provide...

More power to you:

What if you could submit music search queries, right from your mobile phone device, to peer-to-peer networks like Gnutella? These queries would run through your Gnutella client on your home PC, and the files you choose would be downloaded to your home PC. Then you could stream those mp3s over your mobile phone too.

Combine this with song-ID-and-download software; now you've really got something. Imagine this: one afternoon you wander into your favorite cafe and you immediately love the song playing in the background. You tell special software on your mobile phone to identify and grab the song. The software records a few seconds of the song and sends this digital sample to your server or home PC, where software identifies the song, fires up Morpheus (or whatever peer-to-peer software you prefer), runs a search on that song and downloads an mp3 of it. Later, in line at the grocery store, you'd love to hear that song that you caught just a bit of in the cafe. So you plug the headphones into your mobile phone and you stream that new mp3 from your server. Sweet.

Here's how the software would identify the song based on just a sample: it would query the online CDDB database, which contains titles of thousands of songs along with unique checksum signatures for each song. A special algorithm allows software to quickly scan any mp3 and come up with a unique "checksum" or signature, which applies to that song and that song only. So if you have an untitled mp3 on your PC, you can scan it and submit the resulting signature to the CDDB database. The database will spit back the name of the song, if it recognizes it. (I know: the ambient noise would probably screw up the whole scheme, and the checksums in that database probably require -entire- tracks, but can't a guy dream?)

How about a voice interface that lets you quickly request a given song by speaking it into your mobile phone device? Back home, the server software serves the track right away if it's on your hard drive; if not, it tries to find it on Morpheus and streams it down to you as soon as it can.

How about direct streaming from Gnutella users?

I can hardly wait.

[ASCAP and RIAA attorneys: I'm not encouraging anyone to download any music that they haven't purchased. I'm not seeking financial profit from any of this. And I don't have any money. So stay off my back.]

- Sean

Posted by sean at October 26, 2002 05:41 PM | TrackBack (0)

Comments

Sean - thanks for doing some more thinking on this!

A while ago a couple friends and I started getting together to work on a joint design project - not job related, but just something fun. Thing is, we like screwing around more than we liked the product ideas we came up with, so the work never happened.

This seems like an interesting enough project though...you interested in doing a little collaborating with me? Maybe we can come up with some product sketches - maybe even a prototype, if we're really ambitious?

Posted by: Brad Lauster on October 30, 2002 02:03 PM

Brad:

This is a great idea and I'd love to help you bring it to the public.
Thanks for asking.

A disclaimer: between now and New Year's I'll be very busy with grad
school research/applications, with my day job and with a class project.

Given that, I'd love to work on this as time will allow over the next
couple of months, and to devote more time to it after the holidays.

We can start, like you said, by hashing out the idea. We can summarize the
scope of an initial version, look at potential users and platforms,
structure and diagram high-level requirements and data flows, maybe even
put together an early prototype. Sound feasible? I'm at work now, but
I'll get in touch with you and we can discuss this further.

-Sean

Posted by: Sean on October 30, 2002 04:20 PM

This is a brilliant idea, and im sure it'll be very sucssessful.

Congrats

Posted by: Tiago Henriques on November 4, 2002 04:15 AM

Here's proof that the song-identification feature is possible.

The new Neuros digital audio device comes with a built-in FM radio and a feature that will record a 30-second segment of any song on the radio. Back at the desktop, Neuros' "digital fingerprint" feature connects to a Neuros database and provides the song title and artist.

Details:
http://www.neurosaudio.com/store/prod_keyfeat.asp

Posted by: Sean on November 12, 2002 04:29 PM

Posted by: on November 17, 2002 01:05 AM

Posted by: Terry on July 17, 2005 04:10 AM