Somewhere among those 100,000 tracks in your production music catalog is the right one for that TV, film, video, or game. Trouble is, you or your client have to find it. "It's common for people to be surprised at what they have in their catalog," says longtime production music professional Martin Nedvěd. "People can know a few thousand tracks at most, but never the entire catalog."
The composers and production music experts at AIMS API set out to find tracks efficiently. They harnessed the power of machine learning to find the right music fast, using a reference track as a guide. And they did it from a perspective grounded in how the production music economy works.
"Production music has increased in quality in recent years to match and often surpass commercially released tracks," explains Nedvěd. "Now, the biggest problem for music-for-picture professionals is how to find the right track in an endless supply of music when time is always of the essence. We think we have a solution."
AIMS API was built to eliminate these challenges. It recommends and discovers great music options for these professionals in ways that surpass manual search by tags. It can detect similarity quickly at a massive scale and captures the high level of nuance required to serve music supervisors and other sync teams. Its algorithms were trained specifically for production music, focusing the company's efforts on surfacing results that fit briefs.
Nedvěd knows the struggle firsthand, as he spent fifteen years running Studio Fontana, the biggest production music sub-publisher serving clients around Central Europe and managing sub-publishing deals with partners around the world. With AIMS business development lead Einar Helde, he founded an agency to collect neighboring rights royalties for production music performers and libraries in Europe. He's also been running a production music catalog for nearly a decade. Nedvěd knows the challenges rights holders and music teams face, how the pace of production has skyrocketed while budgets shrink. Music supervisors and library owners need to do a lot more with less, and this "more" often involves digging up the right music fast from catalogs with millions of tracks.
The AIMS team saw a great opportunity lurking in the problem, one that machine learning could unlock. Yet it had to yield results that would serve the production music community's existing needs. To do this, AIMS API doesn't simply suggest something similar based on simple audio characteristics like BPM. It hones in on the mood, that certain feel that makes a track work in a particular moment or scene.
"As we were evaluating the results, we were asking if a client sent us this as a reference track, would we send this result back in a playlist?" says Nedvěd. "It doesn't have to be similar in the typical meaning of sounding alike. The right track can be a different genre or tempo, but it has to have the feeling, the use, the vibe."
This foundation in the music for picture and sync ecosystem sets AIMS apart from other music tech companies attempting to build tools for production teams that identify, classify, or search for music using AI. "We initially built AIMS for ourselves, and we've been able to expand our customer base because our results work so well for people in the industry," says Nedvěd. "We're not looking just at surface similarities; we're looking for what makes music work for the industries we serve."