Transformirror is a real-time generative AI interactive video installation developed and created by Daito Manabe and Kyle McDonald, and produced by Rhizomatiks. Beginning with W Osaka, the work has evolved through festivals, city-scale media programs, education events, and new media art exhibitions.
The system combines real-time SDXL Turbo image-to-image transformation with Stable Audio text-to-music. When a visitor is captured by a webcam, their image is immediately transformed and displayed on a large LED screen, creating a public mirror where body, prompt, image, and sound are continuously reinterpreted.
History / Exhibitions and Presentations
2023.12.15 - 2023.12.25 / W OSAKA WET DECK & WET BAR
The original public presentation at W Osaka in Shinsaibashi, Osaka. This exhibition established Transformirror as a real-time AI mirror combining SDXL Turbo image-to-image transformation with Stable Audio text-to-music.
Links: Transformirror Osaka archive
2024.10.24 - 2024.10.27 / KIKK Festival 2024, KIKK in Town
Presented in Namur, Belgium as part of KIKK in Town / the True-False exhibition route. Kyle McDonald's cloud article connects this touring context to GPU logistics, low-latency delivery, and the shift toward browser/cloud deployment.
Links: KIKK in Town feature / Realtime diffusion in the cloud
2025.01.25 / ORGANISM New Media Art Show
Presented at ORGANISM New Media Art Show in Los Angeles. Within a program bringing together artists working across AI, computation, and interactive media, Transformirror was shown as a real-time generative AI mirror experience.
Links: Event information / Kyle McDonald Transformirror History
2025.07.16 - 2025.09.07 / FILE SAO PAULO 2025: SYNTHETIKA
Presented in Sao Paulo, Brazil as part of FILE SAO PAULO 2025. This exhibition placed Transformirror within an international program on synthetic images, generative systems, and contemporary electronic art.
Links: FILE artwork page / Daito Archive: FILE SAO PAULO 2025
2025.10.11 - 2025.10.26 / Cinekid Festival 2025 MediaLab
Shown in Amsterdam in Cinekid's MediaLab program. The festival context emphasizes family and children's encounters with interactive media, making Transformirror a direct, embodied way to experience generative AI.
Links: Cinekid project page / Cinekid program
2025.11.27 - 2026.01.07 / TOKYO NIGHTTIME PROJECT, Shinjuku Neon Walk
Presented outdoors on Shinjuku Chuo-dori as part of TOKYO NIGHTTIME PROJECT / Shinjuku Neon Walk. The installation translated pedestrians into seasonal AI-generated images on a public nighttime street.
Link: TOKYO NIGHTTIME PROJECT
2026.03.27 - 2026.03.29 / The Tech Interactive, AI Literacy Weekend
Presented at The Tech Interactive in San Jose as Transformirror Trickery, a pop-up activity where visitors used props to test how AI interprets the scene.
Link: The Tech Interactive
2026.04.11 / nuitnumerique #23 MUTATIONS, Saint-Ex
Shown at Saint-Ex in Reims, France for nuitnumerique #23 MUTATIONS, described as a real-time generative AI hallucination that transforms visitors' silhouettes into other scenes.
Link: Saint-Ex event page
Related realtime diffusion and body-analysis lineage
Transformirror also belongs to a longer history of camera-based body analysis. Kinect's 2010 consumer depth sensor, together with creative coding environments such as openFrameworks, made it practical to treat visitors as distance, skeleton, and body-part data in realtime exhibitions. Transformirror is not a Kinect-style skeleton tracking work, but it inherits the same shift: the body becomes readable input for a responsive visual system.
Memo Akten's Learning to See is another essential precedent. Its interactive editions used live camera input and neural networks to reinterpret hands and everyday objects as learned visual worlds in realtime. Transformirror extends that question from tabletop objects to a full-body public mirror, using contemporary diffusion models and generated audio.
For technical background, Kyle McDonald's Transformirror History traces related real-time diffusion deployments that grew from the same toolkit. ICC's Generative MV is one such related work, dynamically changing Ryuichi Sakamoto video backgrounds in response to audience text input.
Links: Memo Akten / Learning to See / Microsoft Kinect launch / ICC Generative MV / Kyle McDonald Transformirror History
Reference videos
VIDEO
Myron Krueger - Videoplace, Responsive Environment, 1972-1990s
An origin point for treating the body as camera input and making a visual environment respond to human presence and movement.
VIDEO
Camille Utterback & Romy Achituv - Text Rain, 1999
A canonical camera-based interaction in which the viewer's silhouette intervenes in falling text.
VIDEO
Messa di Voce
A work connecting voice, body, and image in real time, transforming vocal performance into spatial visual expression.
Learning to see: Gloomy Sunday, 2017
An important precedent for rereading live camera reality through a learned visual world in real time.
VIDEO
Daito Manabe + Kyle McDonald + Rhizomatiks - Generative MV at ICC, 2023
A related deployment in which the background of Ryuichi Sakamoto's "Perspective" changed in real time in response to audience text input.
