Halation, physically.
Point a camera loaded with film at a bare bulb at night and the bulb will not stop where it should. A warm red-orange corona spreads out past the edge of the light, fading over a distance far larger than any lens flare would explain. That corona is halation, and understanding where it comes from is the difference between reproducing it and merely adding an orange glow.
The light goes through, and comes back
A frame of film is a stack. On top sits the emulsion — silver halide crystals suspended in gelatin — and underneath it a thick, clear support, usually cellulose triacetate or polyester. Light from the scene passes through the emulsion and exposes it. That is the part everyone accounts for.
What is easy to forget is that the emulsion does not absorb everything. Where a highlight is bright enough, a substantial fraction of the light passes straight through, travels down through the base, hits the back surface, and reflects. The boundary between the film base and the air behind it has a refractive index mismatch, so it behaves like a weak mirror — and at shallow angles, like a very good one. That reflected light travels back up and exposes the emulsion a second time, from underneath.
Because it has crossed the base twice at an angle, it does not land where it started. It lands in a ring around the original highlight, with a radius set by the thickness of the base. Thicker base, wider bleed. This is why halation reads as a property of the stock rather than of the lens: the geometry is inside the film.
Why it is red
Colour negative film carries three sensitised layers, and they are not stacked arbitrarily. Blue-sensitive sits on top, then green, then red at the bottom, closest to the base. Light returning from the base reflection therefore arrives at the red layer first and has to travel furthest to reach the blue one, losing intensity to absorption on the way. The re-exposure is consequently weighted strongly toward red, moderately toward green, and barely at all toward blue.
That is the whole explanation for the colour. Halation is red-orange not because of any aesthetic choice by the manufacturer, but because red sits at the bottom of the sandwich. Halation's default bleed colour is #ff4a1e — heavy red, roughly a third green, almost no blue — which is that layer ordering expressed as a hex triplet.
Manufacturers spent decades trying to kill it
Halation is, unambiguously, a defect. It costs you edge sharpness around every bright object in the frame. Film makers responded with anti-halation layers: a dye or a carbon-black backing applied behind the emulsion that absorbs light before it can reflect, then washes out or bleaches during processing. Almost every modern stock has one.
Which is worth sitting with for a moment. The industry identified the artefact, engineered it away, and then audiences spent the next several decades learning to read its residue as the signature of photographic capture. Cinema stocks — where the anti-halation backing is often less aggressive, and where night exteriors put bright practicals against black — kept enough of it that the association stuck. When a digital image looks too clean, this is frequently one of the things that is missing.
Faking it in five microseconds
The physical account gives a recipe directly. Halation only involves light above a certain intensity, it spreads over a wide radius, it is warm, and it adds to what is already there rather than replacing it. Four properties, four implementation decisions.
First, isolate the highlights. The pipeline drops to quarter resolution and keeps only pixels above a luminance threshold, with a soft ramp rather than a hard cut — a hard cut produces a visible contour where the effect switches on, which no physical process would create. Working at quarter resolution is not only a speed optimisation. Halation is a low-frequency phenomenon; carrying the fine detail through the blur would be wasted work, and the four-tap box filter used during the downsample keeps the extracted highlights from shimmering.
Second, spread it. Two separable Gaussian passes, thirteen taps each, one horizontal and one vertical. Separating a 2D blur into two 1D blurs turns a quadratic amount of work into a linear one, which is the only reason a bleed this wide is affordable in real time.
Third, tint and add. The blurred highlight layer is multiplied by the bleed colour and added to the base image. Addition is the correct operator here and it matters: re-exposure deposits extra density on top of whatever the first exposure recorded. A screen blend or an overlay would lift the shadows too, which is exactly what halation does not do.
Fourth, keep the threshold high. In the Cinematic preset the halation threshold sits at 72 out of 100, in Nocturne at 78. Those are deliberately restrictive values. Halation that touches midtones stops looking like film and starts looking like fog.
Where it earns its keep
The effect does almost nothing on a flat, evenly lit scene, because there is no highlight bright enough to trigger it. It transforms night exteriors, backlit portraits, windows in dim interiors, neon, headlights, and the sun coming through leaves. If you have raised the halation slider and seen no change, the image probably has no specular highlight to bleed — that is the effect being correct, not the effect being broken.
A practical starting point: raise the amount until you can just see the corona at 100% zoom, then back off by about a third. Halation was a flaw that photographers complained about. It should sit at the edge of noticeable, not announce itself.