Back to Work

08

Reimagined with AI

Cover image for Reimagined with AI

About this project

What if the familiar streets of a daily commute could be seen through an entirely different lens one generated live by AI? Could you walk down a real street and watch it become a cyberpunk skyline, or a painterly dreamscape, in real time? This project traces the build of a working prototype that answers that question moving step by step from a desk-bound experiment to a fully mobile, multi-sensory, and interactive AI reality filter.

Reimagined with AI - Live Feed Output

The Build, in Four Stages

1 Proof of Concept (Tethered) A webcam feed into TouchDesigner into Stream Diffusion, restyling live video frame-by-frame based on a text prompt. It worked but it was chained to a desktop.

2 Going Mobile: The Camera Swapped the webcam for an iPhone running NDI HX Camera, linked over a ZeroTier virtual network to a PC. The PC saw it as just another webcam, except now it could walk.

iPhone (NDI HX Camera) → ZeroTier Network → PC (NDI Webcam Tool) → TouchDesigner

3 Going Mobile: The Display Closed the loop by broadcasting the processed video back out over NDI to a relay Mac, then mirroring it wirelessly to an iPad via Sidecar a handheld window into the transformed world.

PC → NDI Stream → Mac → Sidecar → iPad

4 Remote Control via OSC A lightweight control UI on the Mac sent OSC messages over the network to adjust the AI model's parameters (prompt, strength, seed) live, from the mobile rig. Typing "watercolor painting" over "cyberpunk" mid-walk, no laptop required.

+ Generative Audio Google's MusicFX (DJ mode) generated a matching real-time soundscape, routed through a virtual audio patch bay (VB-Audio Matrix) into the same TouchDesigner pipeline and merged into the final NDI output — sight and sound, both AI-generated, both live.

System Diagram

Why It Matters

The most interesting shift wasn't technical, it was in the user's role. What started as a passive "AI filter" pointed at the world became an interactive creative instrument: a slider, a fader, a typed prompt, and the environment itself responds. Not watching AI reimagine reality, directing it.

Links