● ● ●

Ongoing projects 2025-present

01 NEURO-KEYCHAIN

Neuron Keychain is a single simulated nerve cell small enough to carry in a pocket. It runs on a round 1.28-inch screen inside a 3D-printed brain-shaped shell, and it is alive in the only sense a model can be: it integrates whatever current it is given and fires when it reaches threshold.

The cell is stimulated by the body that holds it. A tap is a brief pulse, a hold is a sustained current, a shake adds noise. Each produces a different firing pattern, drawn in real time as a membrane-voltage trace in a 1-bit dither, mint on near-black. Two models run on the device: a leaky integrate-and-fire neuron, one equation and a reset rule, and the Hodgkin-Huxley equations of 1952, four coupled differential equations in which the spike shape, the threshold and the refractory period are not programmed but emerge from sodium and potassium channels opening and closing.

The same cell can be driven by someone else’s brain. A full night of sleep EEG from the open Sleep-EDF database is converted, second by second, into input current and streamed to the keychain over the network. Deep sleep quietens the neuron; REM makes it flicker; the edges of waking make it burst. The device streams its own voltage and spikes back out, so the night, the cell and a projected field of particles move together.

The work is an exercise in scale. Neurotechnology is usually discussed at the level of the whole brain, the headset, the implant. This is the smallest possible version: one cell, one hand, one stranger’s night, and the question of what it means to feel a model respond.

Discipline: physical computing · computational neuroscience · data visualisation · object design

Hardware: ESP32-S3 with 240 x 240 round display, capacitive touch and six-axis IMU (Waveshare ESP32-S3-Touch-LCD-1.28), 1000 mAh LiPo, deep sleep with wake on touch

Models: leaky integrate-and-fire · Hodgkin-Huxley (1952), both ported to run on the microcontroller

Data: Sleep-EDF Database Expanded (PhysioNet), one full night of polysomnography, EEG band powers to input current

Output: on-device membrane trace in 1-bit dither · OSC stream of voltage and spikes · film and live particle visuals in TouchDesigner

Stack: Arduino (C++) · Python, MNE-Python, python-osc · TouchDesigner · CadQuery, Bambu Studio

Enclosure: parametric 3D-printed brain-shaped shell in two battery variants, screen set into the underside, keyring lug

Role: concept and research, neuron models and their embedded port, firmware and interaction design, EEG data pipeline, OSC protocol, visual language, enclosure design and fabrication

Code: github.com/mariatabanelli/neuron-keychain

02 LATENT HEAT

Latent Heat is a subjective archive of the communal energy that gathers in music subculture spaces. It records what photography and video cannot: the collective bodily output of a crowd, the heat, sweat, smoke, breath and movement that a room full of moving bodies generates across a night.

The intensity of a crowd is real, shared and almost entirely unrecorded. Latent Heat captures it as environmental data and gives it back as sound.
A roving archiver logs these atmospheric traces live at real events, and the installation replays that archive as a generative physical soundscape, performed by a field of autonomous electromechanical instruments, each translating one stream of data into a distinct mechanical sound.

The work treats the dancing crowd as a permeable interface between body and environment, and the gathering as something worth archiving on its own terms: speculative, decentralised, communal, and felt rather than filmed.
The archive is not a recording in the usual sense but a record of intensity, the shape of an evening, how the energy of a space rose, held and fell, rendered as something you stand inside rather than something you watch.

Discipline: data sonification · physical computing · installation

Sensing: temperature and humidity (DHT22), carbon monoxide (MQ-7), ethanol (MQ-3)

Output: five autonomous electromechanical instruments and a generative soundscape

Stack: Raspberry Pi · Arduino · Ableton

Role: concept and research, system architecture, sensor and firmware integration, the data-to-sound mapping, the Ableton signal chain, and the industrial design and fabrication of the instruments

Process and development of instruments using servo motors, springs and kinetic repetition through Arduino code.

Exhibition and sound testing.

03 HACKING MACHINES

A fashion collection made using python coding scripts into Clo3D software to dictate the way patterns and drape techniques could form.

By looking into personal memories of sport uniforms and historical research of field hockey and lacrosse, I decided to highlight the way women have re-draped and re-assembled those items of clothing into contemporary form of wearing. Movement from my body was used to generate prints, cuts, patterns and silhouettes.

Discipline: computational fashion · 3D garment simulation · physical computing
Sensing: ody-worn inertial motion sensor (accelerometer/IMU breakout with OLED FeatherWing readout) · rotary potentiometer on Arduino Nano driving the 360 scans in TouchDesigner

Output: a fashion collection of re-draped and re-assembled sport uniform pieces; movement-generated prints, cuts, patterns and silhouettes; 360 body scans animated in real time from a physical controller

Stack: CLO3D (Python API over TCP socket) · Python · TouchDesigner · Arduino · Adafruit Feather · 3D scanning

Role: concept and research (feminist sport history, field hockey and lacrosse uniforms, personal archive), 3D scanning of styled pieces, the CLO3D Python scripting that dictates pattern and drape, sensor and firmware integration, the motion-data-to-garment mapping, the TouchDesigner scan pipeline, and the garment design and construction.

Digital Hyperreallistic Photoshoot

360 full body scans of the models wearing the different looks, then importing the files into Blender and Mixamo to create an environment around them. Nothing about the background, props, decoration, lightning is actually real, only the models and the clothes.

Work in collaboration with Sophie Richardson, BA (Hons) Fashion Communication, class of 2025 @sophiepritchardson

Process & Development:

04 SWEAT-TO-SCENT

Sweat to Scent is a wearable instrument that translates sweat into scent. As the body works, sensors read temperature, humidity, movement and conductivity and interpret them as changing body states: warm-up, endurance, strain, recovery.

A computational layer turns each state into a scent composition, released through a valve system, and a visualisation. Between skin and machine sits a patch of bacterial cellulose, a grown material that absorbs sweat and registers it as visible colour change, so each session leaves a physical trace. Most sports technology turns the body into numbers for measurement and optimisation. This instrument diverts the same signal into smell, colour and material, keeping what is usually discarded: atmosphere, exertion, the sensory and embodied memory of sport.

The work draws on feminist sport history and the idea of the permeable athletic body, the body that leaks, sweats and gives off, and the question of who is held responsible for that.

Discipline: wearable olfactory instrument · biofabrication · physical computing

Sensing: sweat collected at the skin; temperature, humidity, movement and skin conductivity read by the wearable [confirm sensor parts]; a bacterial cellulose patch as a parallel material sensor, absorbing sweat and responding through colour change

Output: scent compositions released through a valve system, real-time visualisations, and a material archive of sweat-marked cellulose patches

Stack: XIAO ESP32-SE · Python · TouchDesigner · 3D-printed valve holder (parametric OpenSCAD) · bacterial cellulose grown and tested at LifeFabs

Role: concept and research, system architecture, sensor and firmware integration, the body-state-to-scent mapping, the valve mechanism and enclosure design, the biomaterial development and testing, and the visual documentation

\

05 MATERIAL FUTURES X NOTHING

An interactive sonic-light installation in collaboration with Material Futures, Teenage Engineering, MidiWare, Nothing and Dropcity.

Using Teenage Engineering and Nothing equipment, the installation became a living instrument, composing a responsive soundtrack to support the growth of mint.

Based on plant bioacoustic research, the sound frequencies act as mechanical stimuli. The work explored how vibration may encourage cell wall growth, nutrient absorption, metabolic activity, and the yield of mint oil.

As visitors interacted with the space, the installation became a living instrument where audience participation had a direct impact on the environment. This miniature ecosystem reflected how even the smallest human actions can contribute to the nurturing of non-human species.

Click press to play

06 KINETIC ECHO: KE-01

Kinetic Echo is a handheld device that listens to how you move and turns it into an evolving audiovisual composition.

Rather than counting steps or scoring performance, it reads the texture of movement, its flow, its bursts, its tension, and composes with it.

It is a small argument that the moving body is a creative instrument, not a set of metrics to optimise.

Year: 2026, ongoing

Role: Concept, hardware, firmware, software, sound design

Type: Handheld device, physical computing, generative audiovisual system

Hardware: RP2350 round-display board, QMI8658 motion sensor, LiPo battery, tactile buttons, haptic motor, custom PCB

Software: Python for motion analysis and the experience signature, Ableton and MIDI, TouchDesigner

Themes: Movement as creative medium, the body as interface, feminist sport theory

Status: Working prototype, in active development

GERBER FILES FOR CUSTOM PCB BOARD

CUSTOM PCB BOARD FOR FINAL PROTOTYPE

HOW IT WORKS:

1 - It reads six qualities of movement

An inertial sensor inside the device tracks acceleration and rotation many times a second. From that raw motion the firmware continuously extracts six qualities:

Flow: how smooth and continuous the movement is
Burst: sudden accelerations, spikes of energy
Tension: held, resisted or restrained motion
Rhythm: regularity and repetition
Drift: slow, gradual change
Recovery: the settling and return to stillness

Each quality is a value that rises and falls with how you move. Together they form a signature: a fingerprint of a gesture rather than a score out of ten.

2 - The screen shows a live bloom

The round display renders a radial bloom, six spokes reaching out from a central hub, one for each quality, each in its own colour. As you move, the spokes stretch and contract in real time, so the screen becomes a live portrait of your movement. The longer and brighter a spoke, the more strongly that quality is present in the moment. Held still, the bloom settles to a quiet resting state.

3 - Capture freezes the signature

A press of the capture button freezes the current bloom. The live portrait becomes a held signature, a single captured moment of movement locked on the screen. This is the hold in capture, hold, replay: the gesture is no longer passing through, it is kept.

4 - Replay plays it back through the body

Pressing replay plays the held signature back as touch. The vibration motor and buzzer pulse in a pattern shaped by the captured qualities, so a burst-heavy gesture feels sharp and staccato while a flowing one feels smooth and sustained. The signature you saw on the screen becomes something you feel in the hand.

5 - Release returns to live

Releasing clears the held signature and returns the device to live mode, following your movement again, ready for the next gesture. Capture, replay and release form the whole loop, and every gesture cycles through it.

6 - The same movement shapes sound

Alongside the screen, the six values stream out of the device over USB into Ableton Live through a small Python bridge. Each quality is mapped to a musical parameter:

Flow to filter cutoff
Burst to note triggers and accent
Tension to reverb and space
Rhythm to delay
Drift to modulation
Recovery to volume and rest

The result is that the same movement that lights the screen also shapes the sound. Motion becomes instrument.

CODE SCRIPTS IN ARDUINO AND THONNY

MAP TO ABLETON FROM DEVICE AND TERMINAL

RENDERED PROTOYPE VARIANTS

07 GLYPH UNBOUND

Nothing's Glyph language lives on the back of the phone.
Glyph Unbound takes that same visual grammar and gives it a body. It is a small round object, built to be held and worn, that renders the identical 25 by 25 dot format the Glyph Matrix uses, so a glyph made for one is a glyph the other could show.

The object has two ways in. You handle it: flip it face down and it falls to a single breathing dot for do not disturb, shake it and it rolls a Magic 8-ball answer, let it rest and it breathes quietly.
Or your phone speaks to it: a notification arrives as a pulse, and a glyph you compose on the phone appears on the screen.
Every change is confirmed by a short tone.

Alongside the object sits a small set of browser tools that send glyphs to the screen over Web Bluetooth.
You can draw a glyph on a grid, draw one in the air with your hand through live tracking, or turn your face into a field of dots and send that.
On the phone they come together as a single installable app, so the whole thing behaves like a product rather than a demo.

Glyph Unbound is an argument that a design language can be extended, not just applied.
It takes Nothing's own grammar and asks what it becomes when it leaves the phone and turns into an object you carry, hold, and read at a glance.

Year: 2026

Role: Concept, hardware, firmware, software, enclosure design

Type: Standalone object, physical computing, dot-matrix light interface, companion web tools

Hardware: ESP32-S3 round-display board, 412 by 412 dot-matrix screen, QMI8658 motion sensor, speaker, LiPo battery, 3D-printed enclosure

Software: Arduino and C++ firmware, Web Bluetooth companion tools, live hand and face tracking, installable web app

Themes: Extending Nothing's Glyph language to its own object, the body as interface, ambient computing

Status: Working prototype, with companion app and public code repository

HOW IT WORKS:

A round dot-matrix screen renders a 25 by 25 Glyph language, the same format as Nothing's Glyph Matrix.

Physical handling drives it: flip face down for do not disturb, shake for the Magic 8-ball, rest for the idle breathing state.

A phone sends notifications and glyphs to the object over Bluetooth.

Companion browser tools let you compose a glyph, draw one in the air with hand tracking, or turn your face into dots, and send it straight to the screen.

PROCESS AND DEVELOPMENT: