The Experience
Step into the room. The Machine pulses, waiting.
In that darkened space, you encounter a volumetric LED matrix, a digital organism dormant until you engage with it.
You are invited to speak into the microphone.
A word, a noise, or a full sentence. Your truth.
Instantly, your voice is physicalised.
It becomes a glowing, moving sphere within the 4 000-pixel volumetric LED matrix.
A digital seed planted in the machine’s memory.
The multilingual system listens, analyses, and categorises what you said, but also how you sounded.
It places your voice into a constellation of previous visitors, retrieving the most closely related sounds from its archive of past inputs.
Suddenly, the room fills with an autonomous, self-reinforcing conversation.
Your voice triggers a cascade of agreement, playing back a sequence of similar thoughts left by others.
The cycle continues.
A loop of affirmation.
Until a new visitor arrives to disrupt the pattern with a fresh sound.
The Concept
Echo Chamber is an interactive installation that explores how voices circulate within closed systems, highlighting the dynamics of algorithmic attention and groupthink. It mirrors the dynamics of social media algorithms and information silos that shape our modern understanding of the world.
In the digital realm, the more we engage with a specific idea, the more we are surrounded by similar voices.
We rarely encounter opposing thought bubbles.
This installation makes that invisible mechanism tangible:
what begins as a single sound becomes an echo of itself, trapped in a growing archive of agreement.
Echo Chamber physicalises this latent space of human language.
By mapping voice embeddings in a multidimensional matrix, calculating Euclidian distances, and visualising the concept into a volumetric LED matrix, we allow visitors to navigate the invisible mathematical distances between their own thoughts and the collective archive of previous participants.
The visualisation in the matrix can be further enforced by the |_| black cube metaphor.
Projected on a wall, you will see cohesive flowers drawing energy from a single |_| Black Cube.
This cube represents the rigid “opinions of the group.” No matter how different the flowers themselves appear, they are all fed by the same singular, unchallenging source.
Technical Specifications
The Echo Chamber combines custom hardware with a sophisticated AI analysis pipeline to create a responsive, generative environment.
Designed for seamless integration into galleries, festivals, and corporate environments.
Hardware & Installation
- The Matrix: A custom volumetric LED screen containing 3 887 pixels.
- Dimensions: 61cm (W) x 90cm (H) x 40cm (D). (Bespoke large-scale versions are available upon commission for site-specific installations.)
- Mounting: The unit is mounted on a stand but is also riggable for suspension.
- Peripherals: Microphone input and audio box.
- Ease of Installation:
- Darkened indoor space.
- Standard power (220V).
- No internet connection necessary (all processing is done local and offline).
Location specific we will foresee:
- Information Radiators: projected slides or printed wall panels with the intro and explanation for the visitors.
- Black Cube metaphor wall projection: projector with image slideshow of magnificent flowers growing from a black box.
Software Pipeline
The installation utilises a hybrid stack of TouchDesigner (Visuals/Control) and a Python Server (AI Analysis) to process everything locally.
The sounds don’t leave the premises.
1. Input & Transcription
- Ingestion: TouchDesigner captures audio and sends OSC messages to the backend.
- Speech-to-Text: Utilises OpenAI Whisper (turbo) for high-accuracy multilingual transcription.
- Translation: Automatic language detection with fallback to English translation (MarianMT) ensures a unified semantic analysis space.
The system is natively multilingual. Visitors can speak in English, Dutch, French or any other language; the AI bridge ensures their ‘truth’ is semantically matched across language barriers, creating a truly global echo chamber.
2. Semantic & Acoustic Analysis
The Machine analyses three distinct vectors to find matches.
- Acoustic Similarity: Uses Resemblyzer VoiceEncoder to map the timbral qualities of the voice.
- Semantic Proximity: Uses Sentence-Transformers (all-MiniLM-L6-v2) to convert text into 384-dimensional embeddings, measuring conceptual closeness via cosine similarity.
- Topical Classification: Uses BART (Zero-Shot Classification) and KeyBERT to categorise the “theme” of the spoken words against a predefined library.
3. Output
- The Machine returns, based on a weighted comparison, a ranked set of echoes (audio files of previous visitors) that match the visitor’s input in tone, topic, and meaning. The best results are played through the audio installation.
- TouchDesigner visualises this internal reasoning and playback in real-time. So, AI is used as a tool to analyse, measure, rank, compare, but everything you hear is the unedited sound of the visitors before you.
Acquire / Exhibit
Exhibition History
- Gallery Ommery (Nov. 2025 – Jan. 2026 | Ghent, Belgium)
During the G.AN.D.A. Art Route. As part of the No Decorum exhibition together with Sydney Fruy, Nico De Nys and Ommery De Zutter. - Wonder Creativity Festival (Oct. – Nov. 2025 | Kortrijk, Belgium)
At Wonder Festival we showed the Echo Chamber as the first part of our Find Your Place exhibition, completed with the panels of our 003 | In Perfect Place series in two rooms of the pre-renovation cohousing project Ten Broele.
More than 15 000 visitors experienced the installation, almost 2 000 sounds were captured.
Bring the Echo Chamber to your space.
This installation is available for loan to festivals, galleries, and events, or for acquisition.
- Exhibition Loan: Full installation setup, including the volumetric brain, microphone station, and location-specific wall-projection components (“The Black Cube” flowers) with technical support (remote or on-site support for installation and calibration).
Designed for both high en low-traffic environments, the installation offers an intimate ‘active’ interaction for the speaker and a captivating ‘passive’ visual/auditory performance for the surrounding audience. - Bespoke Commissions: While the standard unit is available for loan, Sfæ Studio welcomes enquiries for site-specific iterations. We can design and build expanded volumetric screens tailored to the unique dimensions and impact requirements of your venue.
- Commission/Purchase: Bespoke iterations for acquisition.
For technical riders, pricing, and availability, please contact the studio directly.
Or email us directly at [email protected].
Photo Gallery
About Sfæ Studio
Est. 2025 | Heusden/Gent, Belgium
Sfæ Studio is the creative practice of Thomas Willemsen and Michaël Zenner.
Driven by experimentation and a deep curiosity for emerging tools, we blend the physical and digital to create immersive moments that linger. From light design and kinetic objects to AI models and 3D visualisations, our work explores the boundaries of art and technology to tell stories that connect.
We create our own projects, but are also available for commissions.