Home

Old and New Dreams

Old and New Dreams is a large-scale, immersive sound installation in which the movement of visitors through a shared space continuously shapes a generative, computer-driven musical composition. More simply, it can be understood as a kind of symphony in which the instruments — and the way they play — are defined not by performers or a conductor, but by the motion of people in the room.

Crucially, there is a clear link between movement and musical output. Visitors can sense and explore their role in “playing” the piece, as their actions, proximity, and distribution influence the density, phasing, and progression of voices within the score.

The work draws conceptual inspiration from process-based music, while using contemporary sensing technology to transform collective motion into a living, evolving piece. No prior instruction is required; simply being present and moving through the space becomes a form of participation.

TECHNOLOGY

Upon entering, an array of overhead cameras captures your presence. Using a computer vision model — YOLO, short for “You Only Look Once” — the system assigns you an ID, which represents your part in the ensemble. As you move around the room, walking, running, grouping with friends, your ID persists, meaning that if you’re player #5, you remain player #5 for the duration of your visit.

Musically, this is essential, as it enables each visitor to receive a unique melody — and for that melody to follow them as they move.

How that melody may sound varies: the piece itself consists of six sections, each with up to 20 unique parts, ranging from higher-pitched melodies, to lower ostinatos (short, repeating phrases) and bass lines, which take on more of a supportive role. As you enter the space, each visitor is automatically given one of these parts. Critically, the assignment is far from random — rather, the system is constantly taking note of what’s most needed, and working to maintain balance in the ensemble. For example, with people constantly entering and existing, it’s easy to imagine a situation where all of those with melody parts suddenly walk out, leaving too many supporting lines and not enough leads. In such cases, a balancing algorithm prioritizes assigning melody parts to whomever comes in next.

In other words, no matter how many people are present, or in what order they come and go, the ensemble remains balanced overall.

With a small, dispersed audience, each voice remains distinct — a kind of sonic island. Listeners can perceive the larger composition while also hearing their own contribution clearly, and sensing how small movements reshape it

When a small group clusters together, individual voices begin to merge. What were previously separate parts collapse into a single, shared line, emphasizing collective motion over individual action

As more people enter the space and spread out, the number of active voices expands. The music becomes distributed and pointillistic — a field of discrete sounds articulated across the room rather than a single focal gesture

In contrast, large, tightly clustered groups produce dense, interlocking harmonies. Individual voices lose their autonomy, converging into thick, blended textures that privilege mass and resonance over detail