← Back to Projects

Mechanical Meanderings

Mechanical Meanderings is an AI-driven poetic mirror that transforms speech into imagery by manipulating live camera feeds of visitors using their own words.

Mechanical Meanderings — image 1Mechanical Meanderings — image 2Mechanical Meanderings — image 3Mechanical Meanderings — image 4Mechanical Meanderings — image 5Mechanical Meanderings — image 6

About

Mechanical Meanderings is an interactive media installation that transforms speech into AI-generated imagery by manipulating live camera feeds of visitors using their own words. Unlike typical AI art that showcases predetermined aesthetics, this installation functions as a poetic mirror—reflecting both verbal expression and physical presence through algorithmic interpretation. The system operates through a real-time multi-AI pipeline: visitors speak into a microphone, their words are transcribed via speech recognition, reformulated into visual descriptions by a language model, and then transformed into moving images that manipulate their own silhouette captured by the camera. Whether participants engage through poems, storytelling, dance, or spontaneous speech, their expression forms unique and non-reproducible visual feedback that can be read as (human-machine) hybrid subjectivity. The system might not create art per se, but it does invert the usual prompt-response loop, challenging both curatorial authority and self-perception. Rooted in Gordon Pask's cybernetic theories, the installation's design frontloads generative AI models into a 'maverick machine'—a system that doesn't always comply with expectations. Misinterpretations and unexpected visual translations become as significant as successes, embodying the tension between control and novelty that drives human creativity. Participants must actively negotiate with the system's biases, limitations, and quirks, learning to influence its behavior through both voice and movement. By making complex AI accessible through intuitive modalities (voice and body), the installation invites diverse participants to interrogate these systems directly, potentially subverting the norms encoded in training data.

Root Fields

Digital Arts
Storytelling
Phenomenology
Artificial Intelligence