Firekeep
Does the fire keeping loop hold attention for ten minutes without any content?
Firekeep is a physics based fireplace simulator. There are already a couple of games like it and none of them did particularly well, so i’ve decided to try my luck.
Inspiration
During a trip to a lake, i decided to get some fire going because it was genuinely cold, with the wind and all. With a jet lighter that wasn’t too difficult, and i got a fairly hot fireplace going in a couple of minutes. There was something unnaturally soothing in throwing sticks and other combustibles into the flame, blowing on embers beneath to relight the fire and feel the heat, which might make sense if humans did this for thousands of years to survive. That’s roughly where i thought to recreate this feeling in a WebGL demo, and Firekeep got uploaded to itch a month later. How successfully i recreated it i’m yet to figure out.
Design
Given my experience in primarily low-poly modelling and vertex painting, Firekeep models are well, low poly and vertex painted. Cave asset i pasted from AbsurDock, same goes for the skybox (which gets re-baked once on every start, leading to unique cloud composition every run). Flame volume is made from individual animated particle sprite sheets i took from an official Unity blog.
How the Sim Works
The simulation runs decoupled from the visuals, ticking at 8 Hz (dropping to 1 Hz when the WebGL tab is unfocused) while flames flicker and interpolate on top of it with no feedback back into the sim layer.
Logs are chains of discrete segments sharing one rigidbody, with each segment’s fuel and mass derived procedurally from its geometry rather than hand-authored curves. Burn through a segment and the log fractures: the surviving runs split into independent bodies that keep their own heat state.
Heat moves anisotropically: transfer scales with the temperature difference between source and target, but is weighted 5x upward vs 0.2x downward. That’s what makes convection carry heat straight up into a shared flame column while lateral and downward heat falls off sharply (~8 cm half-distance) and only spreads through direct contact, producing the char fronts that creep across a log instead of it burning evenly all at once.
So, did it work?
The question going in was whether feeding a fire for ten minutes, with zero content beyond the loop itself, would actually hold attention. I don’t have outside playtest data yet, no one but me has spent real time with it, so treat this as one biased data point: it holds mine. Whether it holds yours is still open. If you try it, i’d really like to know.