Squarepusher "Terminal Slam" MV

The video imagines a near future where an AR/MR view can rewrite urban scenery: signs and people are detected, masked, erased, replaced, glitched, or camouflaged as part of the music-video image system.

Existing machine-learning tools handled object and human-region detection, while ad positions were completed manually because many urban advertising surfaces change cyclically and remain hard to identify reliably. Those data layers became masks for localized visual processing.

All source images used here were downloaded locally from the Research page and converted to WebP for publication. The frames below keep a consistent ratio so the processing stages can be compared directly.

The glitch system synchronized audio and visual events through a Max patch, exported composition data as JSON, and rendered through a C++ tool. Spleeter was used to separate drums and bass from the stereo track, improving the precision of audio-reactive motion.

The glitch treatment is not just decoration; it acts as a visual language for pulling people, ads, and urban surfaces out of their original layer.

Fake Squarepusher-related advertisements were produced so the city signage could be replaced rather than simply removed.

Credits are organized from the Rhizomatiks work page and the Research BTS page.