The 3D view could not start.
Loading 3D view…
Ready · No fire
Fuel left
O₂ index
Heat index
Fire stage · illustrative
Loading 3D view…
Ready · No fire
Fire stage · illustrative
Precomputed CFAST runs and chamber comparisons. Separate from illustrative Free play.
What this is. An illustrative free-play model with art-directed rendering. Fire growth, spread from item to item, fuel use, oxygen starvation and smoke movement follow simple rules chosen to point the right way. They are dimensionless and use author-set progression, with no conversion to real fire time. A hot smoke layer collects under each ceiling; once it hangs below a door top it spills through open doors and rises up the stairs, so a long, fully involved fire can preheat and ignite fuel in the next room. A shut door only leaks. With every window and exterior door shut the fire uses up the house's air and dies back; opening one lets it grow again, and fire gases are drawn toward that opening. These weights are author-set and sensitive: whether and when a neighbouring room ignites is not a prediction, and no calibrated timing, temperature or fire-modelling result is implied.
What it is not. Not a fire simulation or a prediction. It shows no temperature, gas level, smoke speed, survivability or flashover timing, and nothing here is certified or measured. Model steps are not seconds and do not predict how quickly a real fire develops.
How the picture is made. Soot patterns, char, flames and smoke are drawn from the model's indices. Flame flicker uses the common pool-fire puffing relation (≈1.5/√D Hz) only as an art guide. No fuel is preselected and selecting a room never starts a fire: tapping furniture or choosing from the fuel list only chooses it, and Ignite starts that authored fuel item. Ignite is withheld (Low O₂) when the room's O₂ index is at or below 0.22, the same author-set limit the model uses for item-to-item spread; it is not a measured oxygen threshold. These are illustrative fuel choices, not a calibrated ignition-device simulation. Fresh startup is cold and paused. Reset returns to cold setup, keeping the chosen room, fuel and speed; Ignite then starts a deliberate new run. The see-through side is a single fixed translucent broad viewing face, including the roof slope on that side and the basement below grade, where the ground is drawn cut away as a soil section; the entrance facade, narrow ends, internal walls and street-side roof stay opaque. None of it is a ventilation opening. Camera cutaways only hide surfaces visually; they do not create openings. Doors and windows change the model the moment you tap them; their swing is animation only. Outside, smoke from open and closed openings, eave and ridge seepage, flame and embers at a vented opening, and facade soot follow each room's smoke, burning, oxygen and vented indices. Their shape, wind drift and timing are art direction, not a smoke-flow or wind calculation.
Attic and basement. Same illustrative rules, with a few more author-set ones. The attic is a small space under the roof; its soffit and gable vents are always open, so it is never fully sealed. Smoke reaches it through small ceiling leaks and the hatch; a high opening such as the roof vent lets an attic fire grow much more than a low one. The basement connects to the hall by its stair door. With that door and a basement window open, opening the front door as well draws more smoke and heat up the stairs. These follow the direction of TFS training notes TN106.8 (attic fires), TN106.5 (basement fires) and TN106.4 (flow path), not their numbers. Stored boxes, roof boards and laundry have no measured burn record. There is no water or firefighting yet, and nothing here says when or where to vent a roof.
Readouts. Fuel left is the fraction of the selected room's initial illustrative fuel weights, not kilograms or MJ/m². O₂ index is normalized availability, not oxygen concentration or total air quantity. Heat index is a temperature proxy without °C calibration. Fire stage reports model states: one burning item, intact fuel heating past its author-set ignition index, two or more burning items, or the illustrative flashover state. That state starts only when a room's hot layer and burning are high, it still has air, and an open window or exterior door reaches it directly or through an open doorway; then the layer heats floor-level fuel in that room, and fuel gases the room cannot burn leave through open doors and burn where the next room still has air. Its limits are author-set index values, not the 600 °C gas or 20 kW/m² floor criteria in TN106.4, so it is not a flashover prediction, timing or warning. It never means conditions are safe.
Published numbers for comparison. Not model output.
Primary references. NIST Fire Dynamics explains fuel/ventilation limits and thermal feedback. FSRI Experiment 2 documents a closed structure becoming oxygen limited. These support qualitative concepts; they do not calibrate our coefficients or the cue. No experimental times are transferred to this house. Playback speed changes how quickly model steps advance; 1× is not a real-time fire rate.
Training notes supplied by Shawn. TN105.6 Stack Effect (effective 2019-03-25), pp.2–3: openings, connected vertical paths, height and temperature differences can alter smoke movement, including reverse stack. Our unchanged illustrative transport does not solve wind, pressure or reverse stack. TN105.3 Vertical Ventilation (effective 2011-12-22; revised 2012-01-17), p.3: ventilation is not extinguishment. Its p.16 metal-deck heating example is not a room flashover threshold. These historical notes do not establish current policy or calibrate this model. Roof failure and structural safety are not assessed.
Textures: Poly Haven wood_floor_worn, red_brick_03, asphalt_01 and bark_brown_02, all CC0. Three.js r160, MIT.
Look through a cutaway house, open doors and windows, then choose a fuel and press Ignite to see fire and smoke move.
Illustrative model, not a prediction. Nothing is burning when you enter.