One Hand at the Top of the Scale

Four films, one hand, and nothing in common

Blade Runner 2049, Sicario, Skyfall, and 1917 were photographed by the same cinematographer and finished through the same show LUT. They look nothing alike, and that is not an impression — it is the first thing the measurements say. Their median frames sit 13 L* apart, which is the difference between a film you watch in the dark and a film you watch through fog. One of them leans warm in 87% of its chromatic pixels; another is split almost exactly down the middle.

So this study spends most of its length on the difference. Forty frames, ten from each film, each run climbing from that film's darkest register to its brightest. Then one measurement, at the end, on which all four behave identically — and on which the rest of the measured corpus does not.

What is being tested, and what is only assumed

The shared LUT is the premise, not the finding. Nothing downstream of a delivered Rec.709 encode can confirm which transform was loaded in the DI, and this study does not try. What it can do is ask what a fixed show LUT would predict, and then go and look.

The prediction is specific. A LUT is a fixed mapping, and it has the most authority where it compresses hardest — at the top of the scale, where a wide capture range is being folded into a narrow delivery one. Down in the toe everything is near-neutral anyway; in the midtones, production design, costume and lighting decide. So a shared LUT should show up as scene-independence in the highlights and nowhere else: within a film, the brighter the pixels, the less the scene gets to choose their direction.

That is a claim with a shape, which means it can fail. It does not fail here.

The four films do not overlap

Each film's own 10–90% band on three measures, with its median marked. Same instrument, same units, four different worlds — and this is before anything is claimed about what they share.

How the measurement works

Corpus. Every measured shot of all seven films — 12,582 rows, of which 8,975 are the four. No sampling, no curation: the statistics are the whole population of each film. (1917 is sampled on a clock rather than cut into shots, because it is built as two apparent continuous takes; its rows are samples and it is labelled that way throughout.)

Zones. Each frame's pixels are divided by luma into shadow, midtone and highlight, and each zone reports its own mean a* and b*. A zone must cover at least 5% of the frame (2% for highlights, which are rarer everywhere) before its direction counts at all — the same bar for all seven films.

Direction and agreement. A zone's direction is atan2(b*, a*): which way its colour leans, ignoring how much of it there is. Agreement is the circular resultant R of all a film's shots, weighted by chroma so a near-neutral frame — whose direction is arithmetic noise — cannot vote as loudly as a saturated one. R runs from 0, meaning the film has no preferred direction, to 1, meaning every shot points the same way.

The statistic. Not R itself, which mostly measures how monochrome a film is, but the rise: R at the highlights minus R at the shadows. It asks whether a film's top end is more disciplined than its bottom end, which is exactly the question a fixed transform poses and a production designer does not.

Blade Runner 2049 (2017)

Blade Runner 2049

The darkest and lowest-contrast film of the four, and the only one whose chromatic pixels split almost evenly between warm and cool. The amber sequence everyone remembers is a sequence, not the climate.

Across all 1,901 shots, its median frame reads L* 13.2 with the middle 80% running L* 2.5 to 31.9; mean chroma C* 6.6; 76% of the median frame in shadow; and 49% of its chromatic pixels leaning warm.

Its shadows agree with each other at R 0.405. Its highlights agree at R 0.905, pointing at 97° — a rise of +0.500. The ten frames below climb from the darkest register this film keeps to the brightest.

A figure against a band of amber

Shot 0316 from Blade Runner 2049 at 01:29:38:12. One continuous light trough draws a shallow V across dark concrete; everything above and below falls to near-black, and the leaning figure is placed exactly at the join. — Shot 0316 · 01:29:38:12 · L* 7.7 · C* 11.0 · 86% shadow. Brighter than 29% of the film. Its highlights cover 1% of the frame and point at 78°. One continuous light trough draws a shallow V across dark concrete; everything above and below falls to near-black, and the leaning figure is placed exactly at the join.

A face beside a blown window

Shot 0039 from Blade Runner 2049 at 01:06:11:13. The window is allowed to clip to white while the room around it keeps almost no detail; the glasses, brow and shoulder emerge from that darkness by a narrow margin. — Shot 0039 · 01:06:11:13 · L* 13.4 · C* 3.2 · 83% shadow. Brighter than 50% of the film. Its highlights cover 15% of the frame and point at 111°. The window is allowed to clip to white while the room around it keeps almost no detail; the glasses, brow and shoulder emerge from that darkness by a narrow margin.

Red type across a steaming street

Shot 0160 from Blade Runner 2049 at 01:15:14:12. An illuminated red sign on the vehicle is the only warm element in a blue-black street; steam catches the light from behind and separates the walking figure from the buildings. — Shot 0160 · 01:15:14:12 · L* 14.1 · C* 6.8 · 74% shadow. Brighter than 53% of the film. Its highlights cover 2% of the frame and point at 148°. An illuminated red sign on the vehicle is the only warm element in a blue-black street; steam catches the light from behind and separates the walking figure from the buildings.

A figure between two vertical lights

Shot 0227 from Blade Runner 2049 at 01:20:51:07. Two window slots and a lit doorway supply everything in a green-tinted corridor; the standing figure is lifted just enough to read as a pale blue shape without features. — Shot 0227 · 01:20:51:07 · L* 17.7 · C* 11.6 · 73% shadow. Brighter than 63% of the film. Its highlights cover 4% of the frame and point at 118°. Two window slots and a lit doorway supply everything in a green-tinted corridor; the standing figure is lifted just enough to read as a pale blue shape without features.

White collar against rippling gold

Shot 0326 from Blade Runner 2049 at 01:30:53:12. Reflected water-light crosses the wall in bands of gold, and the white jacket and the face take the same warm cast, so the figure separates by brightness rather than by hue. — Shot 0326 · 01:30:53:12 · L* 21.6 · C* 28.0 · 54% shadow. Brighter than 74% of the film. Its highlights cover 4% of the frame and point at 85°. Reflected water-light crosses the wall in bands of gold, and the white jacket and the face take the same warm cast, so the figure separates by brightness rather than by hue.

A hand with a toy against an open fire

Shot 0516 from Blade Runner 2049 at 01:52:33:05. The furnace mouth burns out to white across the centre while the coat and glove in front of it retain no detail at all, leaving an image built from essentially two values. — Shot 0516 · 01:52:33:05 · L* 29.5 · C* 19.5 · 62% shadow. Brighter than 87% of the film. Its highlights cover 27% of the frame and point at 82°. The furnace mouth burns out to white across the centre while the coat and hand with a toy in front of it retain no detail at all, leaving an image built from essentially two values.

Two men in a flooding cabin

Shot 1863 from Blade Runner 2049 at 03:25:40:02. A single ceiling panel lights the water and both faces from above; the cabin walls and the surface below stay green-black, so the image is built from one reflected highlight. — Shot 1863 · 03:25:40:02 · L* 32.7 · C* 11.1 · 35% shadow. Brighter than 91% of the film. Its highlights cover 11% of the frame and point at 119°. A single ceiling panel lights the water and both faces from above; the cabin walls and the surface below stay green-black, so the image is built from one reflected highlight.

Two men in falling snow

Shot 1892 from Blade Runner 2049 at 03:29:16:01. Snow and pale concrete hold the background at an even light grey; the two dark coats and the fur collar are the only dense masses, and the flakes give the air its texture. — Shot 1892 · 03:29:16:01 · L* 35.5 · C* 7.8 · 28% shadow. Brighter than 93% of the film. Its highlights cover 1% of the frame and point at 163°. Snow and pale concrete hold the background at an even light grey; the two dark coats and the fur collar are the only dense masses, and the flakes give the air its texture.

Bodies behind fogged glass

Shot 0434 from Blade Runner 2049 at 01:44:11:04. Frosted panels at both edges carry pressed hands and a blurred face, and a cold, crowded arcade shows through the gap between them; two figures in dark clothing stand on that boundary. — Shot 0434 · 01:44:11:04 · L* 41.2 · C* 26.0 · 20% shadow. Brighter than 96% of the film. Its highlights cover 12% of the frame and point at 93°. Frosted panels at both edges carry pressed hands and a blurred face, and a cold, crowded arcade shows through the gap between them; two figures in dark clothing stand on that boundary.

A face beneath a grid of white panels

Shot 0359 from Blade Runner 2049 at 01:34:31:08. Lit ceiling panels blow to white and throw a hard graphic lattice behind the head; the pale coat picks up that light while the frame holds almost no colour. — Shot 0359 · 01:34:31:08 · L* 53.8 · C* 9.4 · 29% shadow. Brighter than 99% of the film. Its highlights cover 57% of the frame and point at 100°. Lit ceiling panels blow to white and throw a hard graphic lattice behind the head; the pale coat picks up that light while the frame holds almost no colour.

Sicario (2015)

Sicario

The most decisively warm film of the four, and it spends that warmth from inside a frame that is more than two thirds shadow. Dust does most of the work; the muzzle flashes are too brief to move a median.

Across all 1,205 shots, its median frame reads L* 15.3 with the middle 80% running L* 2.0 to 36.8; mean chroma C* 8.1; 68% of the median frame in shadow; and 87% of its chromatic pixels leaning warm.

Its shadows agree with each other at R 0.618. Its highlights agree at R 0.982, pointing at 98° — a rise of +0.364. The ten frames below climb from the darkest register this film keeps to the brightest.

A face at the edge of the tunnel light

Shot 0931 from Sicario at 02:30:17:16. A single bulb behind her shoulder is all that reaches the frame; the face is modelled by a narrow warm fall-off and the walls disappear entirely. — Shot 0931 · 02:30:17:16 · L* 2.0 · C* 3.0 · 97% shadow. Brighter than 10% of the film. Its highlights cover 1% of the frame and point at 87°. A single bulb behind her shoulder is all that reaches the frame; the face is modelled by a narrow warm fall-off and the walls disappear entirely.

Caged lamps in a smuggling tunnel

Shot 0887 from Sicario at 02:29:52:04. Two caged bulbs and a receding string of lights are the only sources; the earth walls take their amber directly, and the figures in the doorway are almost fully absorbed. — Shot 0887 · 02:29:52:04 · L* 6.0 · C* 9.1 · 92% shadow. Brighter than 25% of the film. Its highlights cover 1% of the frame and point at 86°. Two caged bulbs and a receding string of lights are the only sources; the earth walls take their amber directly, and the figures in the doorway are almost fully absorbed.

Neon behind a bar conversation

Shot 0647 from Sicario at 02:05:55:08. Scattered bar signs throw red, blue and yellow points across a dark room; Kate is lit by a low practical while the man opposite is held as a black mass in the foreground. — Shot 0647 · 02:05:55:08 · L* 9.9 · C* 13.7 · 85% shadow. Brighter than 34% of the film. Its highlights cover 2% of the frame and point at 98°. Scattered bar signs throw red, blue and yellow points across a dark room; Kate is lit by a low practical while the man opposite is held as a black mass in the foreground.

A face turned inside a car

Shot 0346 from Sicario at 01:33:30:04. The car interior is held near black while the windscreen and side windows blow out to the crossing beyond; Kate's face catches just enough of that light to read against it. — Shot 0346 · 01:33:30:04 · L* 15.2 · C* 6.1 · 73% shadow. Brighter than 50% of the film. Its highlights cover 8% of the frame and point at 95°. The car interior is held near black while the windscreen and side windows blow out to the crossing beyond; Kate's face catches just enough of that light to read against it.

A weapon raised past a face

Shot 0376 from Sicario at 01:35:51:15. Shallow focus keeps Kate sharp at right while the barrel and the man behind it stay soft; the red vehicle beyond the window is the frame's only strong colour. — Shot 0376 · 01:35:51:15 · L* 24.1 · C* 9.6 · 48% shadow. Brighter than 71% of the film. Its highlights cover 8% of the frame and point at 86°. Shallow focus keeps Kate sharp at right while the barrel and the man behind it stay soft; the red vehicle beyond the window is the frame's only strong colour.

A face between blown windows

Shot 0062 from Sicario at 01:05:05:09. Two window panels clip to white behind Kate while the interior falls away to near-black; her face and the shoulder of the figure behind her are the only mid-tones present. — Shot 0062 · 01:05:05:09 · L* 27.4 · C* 10.0 · 48% shadow. Brighter than 77% of the film. Its highlights cover 22% of the frame and point at 105°. Two window panels clip to white behind Kate while the interior falls away to near-black; her face and the shoulder of the figure behind her are the only mid-tones present.

Armed men among stopped cars

Shot 0425 from Sicario at 01:37:17:14. Hard midday sun splits the scene into blown road surface and dense shadow; the standing figures are read by silhouette, and the queued traffic supplies the depth. — Shot 0425 · 01:37:17:14 · L* 38.2 · C* 9.9 · 36% shadow. Brighter than 92% of the film. Its highlights cover 32% of the frame and point at 104°. Hard midday sun splits the scene into blown road surface and dense shadow; the standing figures are read by silhouette, and the queued traffic supplies the depth.

Three men in a yellow room

Shot 0454 from Sicario at 01:42:08:00. Fluorescent ceiling panels and yellow walls fill the frame with a single institutional hue; the seated man at centre and the two waiting at left are placed against an otherwise empty box. — Shot 0454 · 01:42:08:00 · L* 41.1 · C* 36.4 · 9% shadow. Brighter than 93% of the film. Its highlights cover 2% of the frame and point at 92°. Fluorescent ceiling panels and yellow walls fill the frame with a single institutional hue; the seated man at centre and the two waiting at left are placed against an otherwise empty box.

A football match beside the fence

Shot 1202 from Sicario at 02:54:32:08. Bright cloud and pale sand fill the frame; the watching families in the foreground and the border fence across the middle distance divide it into three flat bands. — Shot 1202 · 02:54:32:08 · L* 50.2 · C* 16.0 · 10% shadow. Brighter than 98% of the film. Its highlights cover 12% of the frame and point at 107°. Bright cloud and pale sand fill the frame; the watching families in the foreground and the border fence across the middle distance divide it into three flat bands.

A face against an empty sky

Shot 0590 from Sicario at 01:59:42:04. An overcast sky fills the background at a flat pale grey; Kate's face and grey shirt sit only slightly below it, and a distant lamp standard is the only other mark. — Shot 0590 · 01:59:42:04 · L* 56.8 · C* 9.5 · 4% shadow. Brighter than 100% of the film. Its highlights cover 35% of the frame and point at 155°. An overcast sky fills the background at a flat pale grey; Kate's face and grey shirt sit only slightly below it, and a distant lamp standard is the only other mark.

Skyfall (2012)

Skyfall

The most saturated and the highest-contrast of the four, holding both inside a dark frame. It concentrates colour rather than spreading it, which is why so few of its images are remembered as dark.

Across all 2,619 shots, its median frame reads L* 18.9 with the middle 80% running L* 3.7 to 45.2; mean chroma C* 10.4; 63% of the median frame in shadow; and 71% of its chromatic pixels leaning warm.

Its shadows agree with each other at R 0.399. Its highlights agree at R 0.918, pointing at 95° — a rise of +0.519. The ten frames below climb from the darkest register this film keeps to the brightest.

A lit doorway inside a black room

Shot 0002 from Skyfall at 01:01:06:07. Bond's grey shoulder and the small pool of lamplight through the open door carry the whole image; the surrounding wall and banister are held so far down that the frame reads as two lit fragments in black. — Shot 0002 · 01:01:06:07 · L* 1.7 · C* 1.7 · 97% shadow. Brighter than 4% of the film. It carries almost no highlight — nothing here reaches the top of the scale. Bond's grey shoulder and the small pool of lamplight through the open door carry the whole image; the surrounding wall and banister are held so far down that the frame reads as two lit fragments in black.

Gold light across a casino conversation

Shot 1100 from Skyfall at 02:00:54:23. Practical lamps and gilded screens keep the room in one amber register; the black dinner jacket, white shirt and the woman's dark embroidered back are the only strong tonal breaks. — Shot 1100 · 02:00:54:23 · L* 10.5 · C* 16.0 · 85% shadow. Brighter than 28% of the film. Its highlights cover 2% of the frame and point at 85°. Practical lamps and gilded screens keep the room in one amber register; the black dinner jacket, white shirt and the woman's dark embroidered back are the only strong tonal breaks.

Lamplight across a warm flat

Shot 0648 from Skyfall at 01:27:14:20. Table lamps and a chandelier hold the room at a low amber level; M stands at centre in black while an out-of-focus foreground shoulder blocks the left of the frame to near-black. — Shot 0648 · 01:27:14:20 · L* 12.9 · C* 12.5 · 73% shadow. Brighter than 35% of the film. Its highlights cover 1% of the frame and point at 81°. Table lamps and a chandelier hold the room at a low amber level; M stands at centre in black while an out-of-focus foreground shoulder blocks the left of the frame to near-black.

Two men across a panelled office

Shot 2619 from Skyfall at 03:17:31:15. Oak panelling, brass lamps and a naval painting set a warm, evenly lit room; the two dark suits stand at opposite edges of the frame with the desk and window between them. — Shot 2619 · 03:17:31:15 · L* 17.6 · C* 11.2 · 74% shadow. Brighter than 47% of the film. Its highlights cover 9% of the frame and point at 98°. Oak panelling, brass lamps and a naval painting set a warm, evenly lit room; the two dark suits stand at opposite edges of the frame with the desk and window between them.

Olive and glass inside the cell

Shot 1389 from Skyfall at 02:23:10:21. The cell walls tint the image a cool green-grey, against which the olive shirt, pale hair and skin are the only warm elements; dark glass mullions frame the head on both sides. — Shot 1389 · 02:23:10:21 · L* 18.7 · C* 9.6 · 69% shadow. Brighter than 49% of the film. Its highlights cover 7% of the frame and point at 106°. The cell walls tint the image a cool green-grey, against which the olive shirt, pale hair and skin are the only warm elements; dark glass mullions frame the head on both sides.

Flag-draped coffins in a pale hall

Shot 0613 from Skyfall at 01:24:34:10. Receding rows of Union Jacks drive a red, white and blue perspective into the frame; the stone columns and walls stay near-neutral, and a single dark figure stands at the vanishing point. — Shot 0613 · 01:24:34:10 · L* 28.2 · C* 10.6 · 37% shadow. Brighter than 68% of the film. Its highlights cover 4% of the frame and point at 78°. Receding rows of Union Jacks drive a red, white and blue perspective into the frame; the stone columns and walls stay near-neutral, and a single dark figure stands at the vanishing point.

Two silhouettes across a jellyfish screen

Shot 0908 from Skyfall at 01:47:32:01. An animated jellyfish fills the glass wall in cyan and white and reduces both men to flat black cut-outs; the reflection at right doubles the composition. — Shot 0908 · 01:47:32:01 · L* 34.0 · C* 21.8 · 52% shadow. Brighter than 76% of the film. Its highlights cover 25% of the frame and point at 172°. An animated jellyfish fills the glass wall in cyan and white and reduces both men to flat black cut-outs; the reflection at right doubles the composition.

A profile against curtained daylight

Shot 0529 · 01:18:20:15 · L* 43.6 · C* 6.0 · 40% shadow. Brighter than 88% of the film. Its highlights cover 42% of the frame and point at 92°. Tall curtained windows are allowed to burn out behind M, so her white hair and lit profile sit against blank white while her black jacket merges into the unlit foreground desk.

Figures lost in pale smoke

Shot 2314 from Skyfall at 03:01:20:20. Backlit smoke floods the frame to near-white with a cold cast; two figures and a weapon survive as soft grey shapes, and lens flare breaks the light into discs. — Shot 2314 · 03:01:20:20 · L* 67.6 · C* 9.1 · 1% shadow. Brighter than 99% of the film. Its highlights cover 53% of the frame and point at 140°. Backlit smoke floods the frame to near-white with a cold cast; two figures and a weapon survive as soft grey shapes, and lens flare breaks the light into discs.

A vehicle dissolving in dust

Shot 0301 from Skyfall at 01:08:18:01. Raised dust flattens the field to a single pale sand tone; the truck at right and a small dark shape below survive only as soft edges, with very little chroma anywhere. — Shot 0301 · 01:08:18:01 · L* 75.1 · C* 19.0 · 1% shadow. Brighter than 100% of the film. Its highlights cover 98% of the frame and point at 100°. Raised dust flattens the field to a single pale sand tone; the truck at right and a small dark shape below survive only as soft edges, with very little chroma anywhere.

1917 (2019)

1917

The most open film of the four by a wide margin — the least of it in shadow, the most of it leaning warm, and mean chroma still under C* 10. The warmth is a tint across everything rather than colour in anything.

Across all 3,250 samples, its median frame reads L* 26.2 with the middle 80% running L* 2.6 to 54.0; mean chroma C* 9.4; 48% of the median frame in shadow; and 86% of its chromatic pixels leaning warm.

Its shadows agree with each other at R 0.716. Its highlights agree at R 0.930, pointing at 116° — a rise of +0.214. The ten frames below climb from the darkest register this film keeps to the brightest.

Two faces at a lamp underground

Sample 0212 from 1917 at 0:07:02. A small warm source at the right edge is all the light there is; it models two faces and their tunics in orange while the dugout around them keeps no detail at all. — Sample 0212 · 0:07:02 · L* 2.6 · C* 4.9 · 98% shadow. Brighter than 10% of the film. It carries almost no highlight — nothing here reaches the top of the scale. A small warm source at the right edge is all the light there is; it models two faces and their tunics in orange while the dugout around them keeps no detail at all.

Men waiting between the trees

Sample 2732 from 1917 at 1:31:07. Trunks divide the frame into dark verticals and the seated soldiers between them read as one continuous mass; chroma is very low, so the wood is described almost entirely by value. — Sample 2732 · 1:31:07 · L* 12.0 · C* 2.8 · 87% shadow. Brighter than 30% of the film. Its highlights cover 5% of the frame and point at 124°. Trunks divide the frame into dark verticals and the seated soldiers between them read as one continuous mass; chroma is very low, so the wood is described almost entirely by value.

Two figures among burnt trunks

Sample 1052 from 1917 at 0:35:04. Charred trunks break the field into vertical black strokes; the men's khaki and webbing are the warmest notes, and the ground behind them stays grey throughout. — Sample 1052 · 0:35:04 · L* 18.0 · C* 7.0 · 68% shadow. Brighter than 39% of the film. It carries almost no highlight — nothing here reaches the top of the scale. Charred trunks break the field into vertical black strokes; the men's khaki and webbing are the warmest notes, and the ground behind them stays grey throughout.

An archway onto a burning square

Sample 2132 from 1917 at 1:11:06. A black arch frames a square lit only by fire, with a stone cross and a wet street beyond; the man in the opening keeps no detail, and the contrast across the frame is extreme. — Sample 2132 · 1:11:06 · L* 22.0 · C* 15.6 · 69% shadow. Brighter than 44% of the film. Its highlights cover 18% of the frame and point at 85°. A black arch frames a square lit only by fire, with a stone cross and a wet street beyond; the man in the opening keeps no detail, and the contrast across the frame is extreme.

A rifle raised at a ruined house

Sample 1952 from 1917 at 1:05:06. A gutted building stands against pale cloud with its windows as black rectangles; the aiming figure in the foreground is dark, and the whole image runs cool. — Sample 1952 · 1:05:06 · L* 31.2 · C* 6.6 · 51% shadow. Brighter than 57% of the film. Its highlights cover 27% of the frame and point at 146°. A gutted building stands against pale cloud with its windows as black rectangles; the aiming figure in the foreground is dark, and the whole image runs cool.

Wire and mud under a white sky

Sample 0512 from 1917 at 0:17:03. A bleached overcast fills the upper third and flattens the churned ground beneath it; the crouching figures and the tangle of pickets read mainly as dark outline. — Sample 0512 · 0:17:03 · L* 41.4 · C* 9.2 · 28% shadow. Brighter than 73% of the film. Its highlights cover 31% of the frame and point at 119°. A bleached overcast fills the upper third and flattens the churned ground beneath it; the crouching figures and the tangle of pickets read mainly as dark outline.

A body on grass beside the burning barn

Sample 1592 from 1917 at 0:53:05. Green pasture, a burning frame building and standing figures divide the image into three bands; the fallen man in the foreground is the only element at rest. — Sample 1592 · 0:53:05 · L* 42.6 · C* 15.4 · 17% shadow. Brighter than 75% of the film. Its highlights cover 23% of the frame and point at 113°. Green pasture, a burning frame building and standing figures divide the image into three bands; the fallen man in the foreground is the only element at rest.

Aircraft over a green field

Sample 1300 from 1917 at 0:43:21. A foreground back and post are held almost black against a high, soft sky; a walking figure and three distant aeroplanes are the only marks in an otherwise empty upper half. — Sample 1300 · 0:43:21 · L* 48.0 · C* 9.4 · 36% shadow. Brighter than 83% of the film. Its highlights cover 49% of the frame and point at 117°. A foreground back and post are held almost black against a high, soft sky; a walking figure and three distant aeroplanes are the only marks in an otherwise empty upper half.

Chalk walls before the attack

Sample 2852 from 1917 at 1:35:08. Cut chalk fills both sides of the trench and pushes the frame high and cool; the packed soldiers and their khaki are the only warm interruption in a white channel. — Sample 2852 · 1:35:08 · L* 60.2 · C* 11.2 · 5% shadow. Brighter than 96% of the film. Its highlights cover 58% of the frame and point at 167°. Cut chalk fills both sides of the trench and pushes the frame high and cool; the packed soldiers and their khaki are the only warm interruption in a white channel.

A charge seen sideways

Sample 2930 from 1917 at 1:37:44. Ranks of running men cross from right to left while one figure cuts against them; smoke holds the whole image inside a narrow pale band with very little tonal separation. — Sample 2930 · 1:37:44 · L* 72.7 · C* 16.1 · 0% shadow. Brighter than 100% of the film. Its highlights cover 86% of the frame and point at 163°. Ranks of running men cross from right to left while one figure cuts against them; smoke holds the whole image inside a narrow pale band with very little tonal separation.

Where the four stop being different

The one thing all four do the same

Everything above is difference. Here is the exception.

In every one of the four, the shadows are the least disciplined part of the picture and the highlights are the most. The weakest case is 1917, which still tightens by +0.214 — from R 0.716 in the shadows to 0.930 in the highlights. The strongest nearly doubles that. All four finish above R 0.90, which is to say that once a pixel is bright enough, what it is a picture of stops predicting which way its colour leans.

No control film does this. The best of them, Sisu, manages +0.141. Zatoichi goes the other way entirely at −0.376: its highlight direction is very nearly uniform — R 0.092, which is close to no preferred direction at all.

Which way each film's highlights lean

Every film's chroma-weighted mean highlight direction on the CIELAB plane. Arm length is agreement: a long arm means the film's shots point the same way, a short one means they scatter. Amber is the four; steel is the rest of the corpus.

The shadows do as they please; the highlights fall into line

Agreement within each film at three heights on the tone scale. Every one of the four rises and ends above 0.90. Of the three control films, one is flat, one barely moves, and one falls apart entirely.

Would any four films do this?

Four films agreeing is not evidence until you know how easily four films agree. So: seven films divide into a group of four 35 ways. Rank every one of those groups by its weakest member — by how little its least-disciplined film tightens — and see where the actual four lands.

It lands first, at +0.214. The best group that is not the four (Blade Runner 2049, Sicario, Skyfall, and Sisu) reaches only +0.141 — about 66% as much. Under a random split, a group of four would land first 2.9% of the time.

The obvious simpler test does not work, and that is worth stating. Asking merely which way the highlights point puts the four 2nd of 35, not first: their directions sit within 21° of each other, but swapping 1917 for The City of Lost Children gives a tighter group still. A film can arrive at a warm shoulder by being a warm film. What it cannot easily do by accident is arrive there more consistently than it does anything else.

All thirty-five ways to pick four of the seven

Each row is one possible group of four, ranked by its weakest member. The four with the same cinematographer are the group whose weakest member tightens most.

What this cannot separate

One hand, one camera family and one decade are not separable here. All four of these films were captured on Alexa and delivered as Rec.709 between 2012 and 2019. All three controls were not: one is 1995 photochemical, one a 2003 HD master, one a 2022 digital encode. So "the same colourist's LUT", "the same sensor and its highlight roll-off" and "the same delivery era" all predict the result equally well, and this corpus cannot choose between them. The test that would is a film by another cinematographer, shot on the same camera, delivered the same decade — and it is not in this corpus.

The transfer is unratified. Extraction assumed gamma24, a provisional reading of the observed Rec.709 material that the patch-chart acceptance check has not confirmed. Every absolute figure here inherits that. The study is built almost entirely from comparisons within a film and rankings between films, both of which survive a wrong transfer far better than absolute values do.

A fingerprint is not a grading record. None of this recovers a node tree, a mask, a key or an intention. Four films behaving alike at the top of the scale is a pattern in delivered pixels; the explanation is a hypothesis, and the most economical one is not automatically the true one.

A grade you can see, and a grade you cannot

The interesting part is not that a LUT leaves a signature. It is where it leaves one.

Everything a viewer would name as the look of these films — how dark they are, how saturated, whether they feel warm or cold — is exactly where the four disagree most. That is production design, location, weather, lens and the hundred decisions taken before anything reached a grading suite. A show LUT does not touch any of it.

What the four share is invisible in the ordinary sense: a discipline at the top of the scale that no single frame reveals and that only shows up across thousands of them. Which is a fair description of what a LUT is for. It is not there to make the film look a particular way. It is there so that whatever the film looks like, it gets there the same way every time.