2023

Zurich University of the Arts

Focus

Generative images

Physical computing

Main skills

Fullstack development

Python automation

Tsun_Comp

About

TsunComp repurposes a 3D printer into a two-dimensional plotter that visualizes conversations through AI-generated drawings. A custom application interprets dialogue, creates an image, converts it into G-code, and sends it to the printer to bring the exchange to life on paper.

Objective

TsunComp explores how human communication is interpreted, distorted, and reimagined by machines. While we’re used to seeing AI’s responses in text, watching a machine interpret your words and translate them into a real-time drawing creates a more tangible and revealing encounter with algorithmic understanding.

Process

The project began with research into current AI usage patterns, especially around personalization and digital assistance. We then designed a conversational AI that asked light, seemingly harmless questions to build a profile of the visitor. This persona was then used to simulate decisions or behaviors on their behalf within the installation. Great attention and time was paid towards the presentation of the work. It presents itself as an installation featuring multiple layers communicating with each other. The physical installation itself, the web technologies running in the background driving the experience, various electronical components to achieve automation, and a whole branding and identity featuring deliberate design choices.

RESULTS_AND_IMPACT

Project Impact

Users had humurous chats with the AI-personality built into “tsuncomp”. They expressed their feelings, recollected events about their weekend, talked about the weather, and so on. Based on this exchange, they received a “robot-drawn” image that they got to keep for themselves as a souvenir for this exchange.

Personal Impact

While the project concept was pretty straightforward, the main challenges were technological. There were a lot of steps involved, which were cause for headaches and sleepless nights. At the time, the conversion from a PNG image to interpretable G-code commands for the repurposed 3D printer caused many difficulties. Due to this, I was thankfully able to learn a lot about Python automation and also built a backend with it.

GET_IN_TOUCH

andykohl3r@gmail.com

Copy Email