Author: joshua.chelakkattuparambil

The previous blog focused primarily on workshops rooted in neuroscientific research. The following three delve into digital signal processing, visual generation from audio and dynamic spatialisation. Each of these highlights features I am planning to integrate into my personal project.

Moving with time

Moving with Time is an immersive audiovisual performance exploring the evolving relationship between sound, image, and culture through real-time interaction. Amongst all the projects I have seen and been a part of during the entirety of the workshop, this particular performance had the closest semblance to my research project.

Some of the common processes this project shares with the research project include real-time digital signal processing, transforming sound and modifying it in real time. They had used granular and concatenative synthesis in Max/MSP to achieve this. I am still unsure whether to include this in my project, but I would definitely implement an interactive feature for users to stay engaged with the installation. Another common element present in both projects is real-time visual generation, made using TouchDesigner. 

Speaking about the performance, I personally really enjoyed the fusion of music styles from the traditional Persian Setar to the warm textures of the ARP Odyssey and the Monotribe, it progressively builds and resolves eloquently.

Generative Music from the Quantum World

I could have never  imagined using quantum mechanics to make music until Jean-Claude Heudin had talked about it. In this fascinating presentation, we were introduced to ANGELIA, a hybrid generative AI developed to enhance the creativity of the artist for composing, and to augment performative capabilities.

Heudin briefly explained some fundamental concepts of quantum mechanics like superposition, entanglement and measurement (or collapse). He also mentioned how qubits that exist in superposition states, can be subjected to incompatible measurements, and even be entangled with other qubits.

We were also introduced to Qunotes, which are like qubits but with musical notes instead of binary information. The quantum state of a qunote is represented by a linear superposition of a defined number of pitch values. It’s only when the piece of quantum music is played that a qunote collapses into a defined state. Each time the same qunote plays, a different note is obtained depending on the probability amplitudes. This results in many possible interpretations of the same score. Therefore, the composing is done with probability amplitudes instead of fixed notes.

Dynamic Spatial Mixing for Multi-Channel Audio

I was drawn to this project due to our course this semester (Surround Sound and spatialization). I also wanted to know more about spatial rendering and how to implement this in a mixing environment.  This live demo introduces a system for dynamic spatialization and mixing in multi-channel environments using Max/MSP and IRCAM’s SPAT library. The framework combines adaptive spatial rendering with dynamic mixing to create an evolving, responsive sonic field.

It is a unique approach to spatialisation as they combined higher order ambisonics and VBAP to achieve a unified spatial rendering system.The term they used to describe this process is ‘sidechaining’, similar to the one we use in mixing, as audio objects dynamically influence one another and interact with static beds, generating shifting amplitude and frequency relationships and establishing priority-based behaviour between moving and static spatial elements. A parallel binaural rendering was implemented into the system for headphone users as well.

Overall, I found all the sessions at the IRCAM Forum very insightful. There was a diverse mix of topics and workshops hosted by artists, performers, researchers, engineers and so on. I was able to gather a lot of interesting ideas and got introduced to some really cool concepts, some of which I would attempt to apply to my research project.

Last month, I had the opportunity to attend the IRCAM Forum in Paris with my classmates. It was very enjoyable and provided a pathway to explore a variety of projects in the innovative fields of sound technology and research. The workshops were generally well presented and featured discussions that helped attendees better understand the concepts and workflow. Some of these projects really stood out to me, and I would like to share them in detail throughout this set of blog posts.

Liberated frequencies 

“Liberated frequencies” is a Bio-interactive Audiovisual performance by Keigo Yoshida which combines neuroscientific research with machine learning to generate sounds in real-time. During this performance, an AI is trained to learn favorable patterns and generate sounds from signals triggered by the user’s EEG sensors. In other words, the brainwaves (theta waves in particular) of the user are measured and sonified to form abstract compositions from preferred sound textures.

The team behind the installation collected a library of glitch, experimental, voice and noise sounds, which the subject rated based on the pleasure they evoked. During the performance, the AI learns in real-time from the most preferred sounds. However, instead of simply creating a comforting environment, the AI utilizes this data to intentionally disrupt the experience. Once high auditory pleasure is detected, the system alters the pitch, tempo, and rhythm, pulling the listener out of their comfort zone to challenge human-centric aesthetics and create an unpredictable, evolving soundscape.

The RAVE (Realtime Audio Variational autoEncoder) framework was used to process and generate the sounds, while the visual representations were made and mapped through TouchDesigner. The translation of data to an audiovisual form in real-time is one of the primary aspects of my research for my personal project. Therefore, I remain open to explore more in the world of real-time data audification and visualisation. This brings me to the next workshop, which is also based on neuroscientific studies.

Brainwave-Modulated Generative Music

This live demo presentation was done by Zap Bain, musician, sound engineer researching on auditory neurofeedback and spatial audio. This project also dwells on real-time EEG brainwave control of spatial audio systems. Despite not being familiar with neuroscience, I found the research very fascinating and it made me curious about neurofeedback responses in general.

The artist used embodied cognition techniques including movement, neural asymmetry, vestibular system engagement and brainwave frequency ratios for this research. The presentation showcases ambisonic work developed with SPAES using Spat~ and boids algorithms.

I was able to link some of the methodology behind this presentation to my own research project. I was exploring ways to utilize sounds in a surround sound/ambisonic setup just using vocal feedback and this session gave me some hints on how I could integrate different parameters to control the sound signals and how it could be linked with the visual element to make it more understandable.

Evolving the Living Looper: Artistic Research, Online Learning, and Tentacle Pendula

The Living Looper: Rethinking the Musical Loop as a Machine Action-Perception Loop

I came across the concept of ‘living loops’ while going through the NIME papers. Naturally intrigued, I wanted to learn more about this interesting topic. It is a special music interaction project that can yield vastly different outputs, mainly due to the use of generative machine learning models. The implementation focuses on the RAVE encoder, created by IRCAM, where the audio data is mapped into a compressed latent space and trained using an autoregressive predictive model incorporating partial least squares regression. The dataset primarily focuses on the timbre of the sound rather than the overall musical structure.

It is a unique approach to looping as the artist has the ability to control the loops and train them for real-time audio synthesis, which help create more unique textures and tones. Though the instruments used for the interface were primarily stringed and wind instruments like the violin and saxophone, I wonder how the results would vary if percussion was used. Was it possibly avoided due to the fast attack and low sustain times? A thought to ponder, for sure. I am also curious to learn how turntablists and DJs could use this tool in their performances.

The 2025 paper introduces a user-centric improvement to the previous model, which lacked a visual interface. The upgrade also features a visualisation for each loop by introducing a unique ‘tentacle pendulum’ that interacts when the loop is played. The design was intended to represent the RAVE latent dimensions in order of importance, where the value of each latent dimension determines angular displacement and hue of each segment. The introduction of incremental algorithms which distribute the computation across audio frames, potentially allows more computational resources to be brought to bear on each loop. 

One of the other things I really like about this project is its compatibility and open-source accessibility. Any model following the RAVE API can now be packaged into a Living Looper instance using a Python CLI available on PyPI. Since these instances are now themselves nn~ models, the core functions can be loaded directly into Pure Data and Max, and the new graphical version can be easily installed as a SuperCollider extension, making it very accessible for most sound and music enthusiasts.

One issue I assume they would face is latency from the signal processing, making the timbre of the sound a crucial indicator for the looper. The parallel processing of effects like delays and reverb could prove to be quite a challenge as well, and I would be interested to know how they could overcome this challenge.

Overall, I thought it was a super interesting project and especially useful for artists and musicians to experiment, compose and express themselves. Excited to see what the next follow-up to this project is.

On the final day of the Elevate Festival (March 7th, 2026), I had the privilege of attending a workshop led by the renowned composer, performer, and turntablist Mariam Rezaei, alongside my fellow classmates from the Sound Design program. I wanted to write a short blog post about my experience during the session and the insights I gained into this intriguing art form.

Based in the UK, Mariam Rezaei has built an astounding repertoire through her diverse range of performances and compositions, collaborating with artists from all over the world. She has such a unique approach to turntablism and was kind enough to share her insights on the interesting world of vinyl players and record spinning.

I went into the workshop not knowing exactly what to expect and feeling like I knew far too little about the performative art of turntablism. Some of the performances Mariam did were so distinct that they mimicked actual acoustic instruments with total authenticity. This was made apparent when she played a sample of a Japanese wind instrument (likely the shakuhachi). The velocity, tempo, and pitch were all controlled meticulously using only the vinyl player. It became clear that beat matching and pitch matching are second nature to her, developed through an exceptionally trained ear from years of experience.

The immense dedication she has poured into her craft was evident throughout her performances; everything from the hand-eye coordination to her perception of pitch and time was outstanding. The ability to seamlessly switch between tracks and sync them perfectly without breaking a sweat is a skill that is criminally underrated and underappreciated by the general public.

Speaking of which, she briefly discussed the bias turntablists face while performing in orchestras and philharmonics, as many critics still do not classify the turntable as a real instrument. However, she is gradually helping to break this stereotype by demonstrating how valuable and resourceful the turntable is, especially when recreating the textures of specific instruments. She described the experience as blending so seamlessly into an arrangement that listeners are unable to tell if the physical instrument is actually present on stage.

It was also intriguing to see her building on established styles and redefining them as her own. This was evident in the different techniques she demonstrated, from beat juggling to fractured beats; each technique came with her own signature “spin” (no pun intended). She showed us that turntablism isn’t just about playing records, but about deconstructing and reassembling sounds in real-time in the most artistic way.

Overall, the workshop broadened my view of turntablism. It is a demanding, nuanced art form that deserves far more recognition than it currently receives.