Meet the programmers: Accusonus

Is it possible to mix artificial intelligence with audio software development? Alex Tsilfidis reckon so…

We fire a salvo of questions at Alex Tsilfidis, one half of Accusonus, the development team behind a string of plugin hits including the mix-separating Regroover, mic bleed-eliminating Drumaton, and noise- and reverb-reducing ERA-D.

How did Accusonus get started, and how would you describe the company’s philosophy?

“Accusonus was founded by Elias Kokkinis [left] and myself. As undergraduate students in Electrical and Computer Engineering, we were both passionate about music and music technology. After graduating, we pursued PhDs in Audio Signal Processing and ended up in the same lab. We founded Accusonus right after, with the vision to use science to improve the everyday lives of musicians, producers and sound engineers.

“There are three principles that define the Accusonus culture. First, we build artificial intelligence tools to empower music makers, not to replace them. Second, music has many complex dimensions and music technology has to take them all into account. Most consoles and DAWs are built around the idea that processing several tracks individually and then summing them is all we need - we believe it’s not! Finally, in traditional sound engineering, one sees the world as a collection of microphones and tracks - at Accusonus, we see the world as a collection of sound sources.”

You’ve tackled mic bleed in Drumatom, and reverb/noise in ERA-D. Do the processes at work in each largely match up?

“They are both complex products and really hard to create, since we had to invent new algorithms for both of them. Although the algorithms are different, there’s a core similarity. Both products are inherently multichannel: Drumatom works on a multichannel drum kit, while ERA-D takes advantage of secondary microphones to improve the sound of the main microphone.”

Unmixing’ is often called the holy grail of music production. So with Regroover, how close do you think you’ve got?

“Sound layers for Regroover are not necessarily ‘unmixed’ sources, but sound elements with musically meaningful temporal and spectral characteristics. The most inspiring Regroover applications we’ve seen so far came from artists who use it to produce new sounds and not to just access old sounds that were already there.”

Your products must be pretty complicated under the hood… but how do you choose which parameters the user gets access to?

“Our products our powered by new-generation algorithms with many, many parameters… but these parameters don’t necessarily make sense in a musical context. The question is, how can we map them into meaningful user controls without messing the workflow? Our answer is to build simple-to-use products that enable fast results and at the same time offer a degree of advanced functionality for cases where time is not an issue.”

What’s next for Accusonus?

“This summer we raised a round of investment to help us grow the company. We already have a brilliant team in place, and now more talented people are joining the company. We’ve received incredible user feedback for Regroover, and we plan to integrate some of these great ideas into the product very soon. Our research team is working on new AI and deep learning technologies, and I’m very excited about the potential!”  

Get over 70 FREE plugin instruments and effects…
…with the latest issue of Computer Music magazine