At the recent DAFx26 conference, iZotope’s own Principal Research Engineer, Alexey Lukin, took the stage to present a masterclass on spectrogram-based audio processing and restoration.
Alexey, who holds a Ph.D. in Applied Math and Computer Science, has been a driving force behind iZotope’s core digital signal processing (DSP) technologies since 2002. Over his long tenure, his pioneering work has helped shape flagship innovations ranging from the IRC maximizers in Ozone to the advanced algorithms powering iZotope RX, iZotope Radius, and MBIT+ dither.
iZotope Principal Research Engineer Alexey Lukin presenting the keynote at DAFx26In his presentation Spectrogram-Based Audio Processing and Restoration, Alexey unpacked how frequency visualization aligns with human psychoacoustics and traced the history of noise reduction from early hardware gates to modern neural networks. Highlights included a deep dive into Short-Time Fourier Transform (STFT) windowing tradeoffs and an insider look at iZotope's machine learning stack.
Keynote attendeesAttendees got a look at how RX has evolved from its first machine learning modules in 2017 to current state-of-the-art Band-Split RNN and BS-RoFormer architectures, showcasing the impressive quality jumps in tools like Dialogue Isolate across RX versions.
While the conference was an in-person event only, you can hear from Alexey about RX in our recent BorisFX livestream, where he dives into the neural net improvements in some of RX 12’s most popular tools. Click here to jump to that part!
