The Evolution of iZotope RX according to Principal DSP Engineer Alexey Lukin

Having been used to repair audio on hundreds of blockbuster films over the years, iZotope’s acclaimed RX restoration and post production dialogue editing suite has firmly established itself as the industry standard in the modern movie-making business. Headliner recently caught up with the company’s principal DSP engineer, Alexey Lukin, to find out how RX really works, what makes it so popular, and what to expect next...

RX has long been the go-to audio repair and polishing suite for film, television, music, podcasts, video games, sample libraries, and more. In a quest to find out how a complex set of algorithms could be turned into one of the most widely used, versatile tools in the business, Headliner spoke to iZotope’s principal DSP engineer and one of the original brains behind RX, Alexey Lukin.

Headliner starts by congratulating him on the recent news that iZotope had been awarded with a scientific and engineering award by the Academy of Motion Picture Arts and Sciences. A rather incredible achievement to say the least, and although RX7 was previously recognised with an engineering Emmy a few years back, this latest achievement almost feels like the icing on the cake. Lukin describes it as “a great honour.”

It’s fair to say that Lukin has witnessed the full evolution of RX and seen it develop into a truly game changing industry tool, as well as seeing iZotope grow significantly as a company:

“RX has a long history indeed, and when we first began working on it in 2003, we were pretty much just a five person company,” he reveals. “Then in 2007 when we released it, we were probably a couple dozen people, but now we’re more like 200.”

Lukin started working at iZotope from Moscow around 2002, initially on a part time basis while he was still doing his PhD studies. Once he got his degree, he moved to Boston in 2006 to work for the company full time.

“Since my youth I’d been very interested in music, and then I became interested in music gear and audio quality. I was just trying to get my noise levels down, you know?” he chuckles. “Later at university when I started studying digital signal processing, I began to contemplate DSP that could help me with noise reduction.

“It was probably around 2003 when I had those first ideas, and I developed some of the algorithms that eventually went into RX in my PhD thesis. It was all about how signal processing is related to how we perceive sound, so essentially using psychoacoustic models to improve the quality of audio processing algorithms. These first ideas that later became RX were bubbling even before I joined iZotope, and of course RX has definitely been in good hands since then!”

Before he joined the company — which back then was known mainly for plugins like Vinyl and Ozone — Lukin was prototyping some early noise reduction algorithms and coding in C++ mostly for standalone applications. His implementation of these algorithms was soon poured into some of iZotope’s early plugins, that unfortunately never saw the light of day:

“After that we understood that in noise reduction, the real time workflow of a plugin isn’t always the best choice; you may want to look at your sound more thoroughly offline, find the events such as chair squeaks or dog barking or footsteps, and be able to select and edit them precisely. Plugins do not really allow for this kind of workflow, so it was around 2005 when we realised that we want to work with the spectrogram.”

Despite it not being the first time a spectrogram had appeared in an audio editor, Lukin believes iZotope RX was the first application that actually combined waveform and spectrogram on top of one another, so that sound engineers were able to see the familiar representation, the waveform, but also the more deep representation at the same time, the spectrogram.

Introducing this in RX enhanced the versatility of its sound editing capabilities indefinitely, with the adaptive spectrogram enabling users to see more detail in their sounds than ever before.

The first version of RX, despite having all the fancy spectrogram displays, only had “very primitive” selection tools, and only allowed a rectangular time frequency selection. It also only had five modules: Hum Removal, Denoise, De-click, De-clip and Spectral Repair, “but even that was a good breakthrough because of the adaptive spectrogram and also being able to apply the tools to arbitrary time frequency selections,” Lukin points out.

“Interestingly, the first version of RX was not as heavily steered towards audio post production, and we actually didn’t understand the market very well back then. We hoped that people would be using it in home studios, such as for podcasting, or for the restoration of old records, or even for things like forensics.

“Over the years we received a tremendous amount of feedback, and we formed some amazing relationships with post production professionals. We’d always taken that feedback very seriously, and we realised that there was so much we could do to speed up workflows and do some new, previously impossible kinds of audio processing for these people. That’s why, a few years later, we began really targeting dialogue editors.”

More recently, iZotope has sought to make RX a more interesting proposition for music production professionals, with modules like Guitar Repair and Music Rebalance being of particular appeal to producers, DJs, remixers, and any audio engineers working with instrumental vocal recordings.

The latest version of RX has three sub-modules within the Guitar Repair module. One of which is designed to eliminate fret squeaks or string noises, another to eliminate excessive picking or plucking of guitar strings, and the third to attenuate buzzing or humming of an amplifier.

Another feature aimed at music professionals is Wow & Flutter, designed to attenuate speed variations and pitch fluctuations in old analogue media such as tape or vinyl. Then there is Spectral Recovery, which allows users to recreate missing high frequencies of sounds that have been low pass filtered, such as those going through a lossy encoder like Zoom or Skype. iZotope also added a few improvements to existing modules like Music Rebalance, as well as overhauling RX’s batch processing capabilities.

“Both the Guitar Repair and Music Rebalance modules have received considerable attention,” Lukin shares. “Wow & Flutter is more specialised but it’s appealing to a narrow group of professionals who are restoring old recordings, and we’ve been seeing some good traction and feedback on that.”