A laser wall that reacts to a room using a local language model
Claude Code built the speech, motion and decision loop driving a laser projection at an art residence in Vilnius.
Credits
Earliest dated source:
The full story
The poster, who does not write code, vibe-coded this with Claude Code for a free art event at the ZOLAK residence in Vilnius. Microphones feed live speech-to-text, cameras feed person and motion detection on a Coral Edge TPU, and a local LLM on an RTX 3090 turns that into what the laser wall shows next. Each job runs on its own device so perception does not compete with the LLM for VRAM.
A 78-second test showed spoken words coming back as huge projected type a moment later, and people began testing it and talking to the wall.
How it was done
Tools
Themes
Loading themes…
Sources
From a short post by the maker, so the details are as they tell them.
- Reddit: I can't code, but I vibe-coded a live AI party installation with Claude Code: a local LLM listens toReddit poster: vanadiuzPosted
The maker's post on r/ClaudeAI.
If it disappears: archived copy of Reddit: I can't code, but I vibe-coded a live AI party installation with Claude Code: a local LLM listens to (opens in a new tab)