Back to  Projects

WHEN  EMOTION  BECOMES  DATA
Multimodal installation,  2026



Warning: Stroboscopic lighting, intense sound and photosensitivity risk.








Role                  Creative Technologist
Format           Multimodal Installation
Tools                TouchDesigner,Arduino,GPT-2
Year                  2026


Exploring the profound gap between human emotion and AI comprehension, this multimodal installation translates subjective feelings into real-time generative audio visuals. Powered by custom language models, it compresses and translates nuanced human text to construct a highly biased algorithmic world, critically exposing the machine's inherent distortion of emotional reality.

TRANSLATION

Using GPT-2, the system intentionally exaggerates the human machine cognitive gap by reducing complex emotions to rigid "keywords." By selecting these keywords via a custom UI, users trigger real time algorithmic narratives and visual feedback on the LED screens.






OUTPUT

The installation utilizes a dual LED screen architecture. The lower monitor outputs machine generated, distorted textual narratives, while the upper monitor displays algorithmically reconstructed human silhouettes.

Anxiety Modes: System parameters trigger aggressive red text distortion, visual fragmentation, and broken typographical connections.
Anxiety Modes: System parameters trigger aggressive red text distortion, visual fragmentation, and broken typographical connections.
Anxiety Modes: System parameters trigger aggressive red text distortion, visual fragmentation, and broken typographical connections.
Anxiety Modes: System parameters trigger aggressive red text distortion, visual fragmentation, and broken typographical connections.
Dissociation Modes: Parameters induce extreme blur, motion delay, and corrupted signal effects resembling broadcast failure.
Dissociation Modes: Parameters induce extreme blur, motion delay, and corrupted signal effects resembling broadcast failure.
Dissociation Modes: Parameters induce extreme blur, motion delay, and corrupted signal effects resembling broadcast failure.
Dissociation Modes: Parameters induce extreme blur, motion delay, and corrupted signal effects resembling broadcast failure.
System Overload: Upon detecting discrepancies between textual and visual states, the system triggers high intensity stroboscopic white light and auditory alarms, physically simulating algorithmic collapse.
System Overload: Upon detecting discrepancies between textual and visual states, the system triggers high intensity stroboscopic white light and auditory alarms, physically simulating algorithmic collapse.
System Overload: Upon detecting discrepancies between textual and visual states, the system triggers high intensity stroboscopic white light and auditory alarms, physically simulating algorithmic collapse.











SYSTEM BUILD
01 Network & Input
Local server connects the UI with TouchDesigner and captures user-selected keywords.
02 Core Processing
TouchDesigner processes text input in real time and generates dynamic visuals.
03 Display Routing
Visual output is sent through video processors to the physical LED matrix.
04 Physical Feedback
TouchDesigner sends a serial signal to Arduino to trigger stroboscopic lighting when an intentional system error occurs.


Back to  Projects