← all posts site ↗

dioramas, not sprites, for the cctv game

the premise is a killer slowly getting closer and all you have is a map of cctv cameras across the roads. most are down, some are up, some are too dark. the original plan was do it like 1993 — flat sprites, a few facings, pixel art style. no. a cctv feed needs dimensionality. it can be low resolution with grain and dither and a choppy framerate, but that is all aesthetics — the view still has to be a real camera's view, and those views vary a lot.

dioramas win. small blocks, virtual cameras, toggleable views, and one big consequence: the car is actually driving the road graph in real time, so prediction becomes real. he was at x heading north, so he could be at y or z — that becomes a true statement about a simulated car, not a scripted illusion. three dimensional here is not about looking fancy, it is about the simulation being honest.

the architecture stays cheap on purpose: one road graph, one car entity, waypoint driving, and each camera is just a small render texture that only renders while its window is open. close the window, rendering stops, but the simulation keeps running — reopen the camera and the car is mid frame exactly where it should be. twelve dioramas never all render at once, so it stays fast even on modest hardware.