December 17, 2010 at 11:33 am

Project: Record a Song with uJam’s Artificially Intelligent Accompaniment

Six months after a start-up called uJam took second place at the TechCrunch Disrupt conference in New York last May, the company’s artificially-intelligent recording studio — deemed one of the neatest mash-ups of music and artificial intelligence by Wired.com earlier this year — emerged from private alpha on Friday. Now, anyone can use it to turn their warblings into full songs. (Listen to my somewhat embarrassing effort below).

“Oh,” you say, “but I don’t know how to sing.” What do you think computers are for?

In part, they’re for making your voice sound better and fleshing out your melody with smart accompaniment with drums, guitar, keyboards, and so on, in little more time than it takes you to sing the melody. If you can play an instrument, uJam can build a song around that, too. Suddenly, your lonely melody for recorder is transformed into a full, harmonically-textured masterpiece. Sort of.

To give you an idea of what is and isn’t possible with uJam, I recorded a little ditty this morning about Evolver.fm, which also officially launched today. (If you’d like to give uJam a try, you might want to create a membership on the site soon; only 777 invitations were still available as of this writing.)

Evolver.fm’s uJam song

I made that in about three minutes, even though it’s been a few months since I’ve used the private uJam alpha, and I’d forgotten my way around the controls. Not only was it quick, but it was easy.

Once you've recorded your base sample with voice or instrument, uJam, which launched to the public on Friday, lets you add specific instruments to the mix as well as a general backing track.

After selecting the clicktrack option, I sang the first words and melody that came into my head, aligned my singing with the beat, chose from a wide selection of backing tracks, vocal effects and autotune percentages, and that was pretty much it. Throughout the process, uJam’s Flash-programmed interface worked smoothly, and all of the controls are self-explanatory.

According to the company, real musicians recorded these backing tracks, which can be tweaked in all sorts of ways for your interactive pleasure.

“UJam’s proprietary ‘Song DNA’ technology interprets musical input through an intuitive user interface, then transforms it into a composition and renders it with real, specially prepared audio tracks that have been produced in professional studios around the world,” wrote a company spokesman. “That means studio-quality results for everyone who uses uJam — and you don’t have to be an engineer to figure it out.”

The program also permits more granular tinkering, allowing you to alter the instruments within a backing track, add new instruments to the mix, and change pitch and tempo. Given the fact that all of these relatively sophisticated functions require no experience with computer audio recording, or even with singing, the results are fairly impressive. And for everything I just described with vocals, equivalent features exist for using an instrument as the originating melody. A free account (the service still claims to be in alpha) allows users to create up to ten songs in this fashion.

UJam is even more impressive now than at its debut, when it wowed the TechCrunch crowd. Upcoming features include song sharing (so you don’t have to wrangle the MP3 on your own), a song structure feature that will let you sculpt passages, and artificially-intelligent melody creation that obviates the need to provide the initial kernel of melody, if you don’t want to come up with that on your own.

When uJam starts, all you hear is a click track — just like Prince, alone, in the studio.

When you’re finished, minutes later, you’ll have an MP3 suitable for emailing, posting on your blog, sending to friends, and so on. If we all communicated with uJam songs instead of email, the world would be a more interesting — if inefficient — place.

See Also: