Grade reconstruction report · 2026

Small Pools in a Black Room

Eight frames from the dark passages of The Bride!, a film whose median frame sits at L* 6.7, where most of the picture is held at black on purpose and every face, lamp, spotlight and neon sign has to be kept alive in a small pool above a floor that an 8-bit delivery can barely describe.

The Bride! is one of the darkest films this repository has measured. Across its 2,324 photographed shots the mean L* is 9.23 and the median frame sits at 6.7; the shadow-field share averages 0.83. It gives most of its frames black, and places small, specific lights in it: a table lamp, a fringed shade, a blue follow-spot, neon out of focus, streetlamps in haze, a gas-station sign, a lit tower.

That changes what a grade reconstruction has to explain. The palette is not the hard part; each frame carries two or three colours at most, and they come from the lights in shot. The hard part is the floor. Where exactly is black, how much of the picture sits on it, what is kept one step above it, and how does a face survive at L* 20 inside a frame whose average is 5? Every study here has a deep-shadow share between 0.56 and 0.79 — more than half the frame within a few code values of black.

The eight studies cover the problems that follow from that: a monochrome face rising out of black, a kiss beside one small lamp, a man turning from a fringed shade, dancers in a single blue spotlight, a masked face against defocused neon, two figures walking into haze and streetlamps, a gas station at dusk under orange neon, and a lit tower that is almost certainly built.

Subject
The Bride!
Studies
8 frames
Measured against
2324 eligible midpoint frames
Interpretation
gamma24

The film-level construction

What holds the look together

Black is the majority of the frame

Every study here puts between 56% and 79% of the picture in deep shadow; across the film the median frame sits at L* 6.7. This is not an underexposed look lifted in the grade; it is a picture designed to be mostly black, with the grade's job being to keep that black clean and to decide what is allowed to rise above it.

Light arrives in small pools

The highlight share of every study is at most 2%. Brightness comes from small sources and the few surfaces they touch: a lampshade, a spotlight cone, a streetlamp, a sign. The pools are not graded up to fill the room; they are held so that their falloff into black stays visible.

The dark has a colour; the sources have another

The ambient darkness usually leans cool — teal-green in the street and the tower, blue under the spotlight — while the sources are warm, rose, orange or neon. Mean chroma stays low (C* 0.5 to 11.3 in these studies) because most pixels are nearly black, yet the palettes show two families each time. Keeping the two in separate branches is what stops a warm lamp from warming the whole room.

Faces are the one thing lifted

Where a face appears it is the brightest broad surface in the frame after the practicals: the monochrome face out of black, the blonde hair beside the lamp, the man turning from the fringed shade. Face work lifts eyes and the lit plane a small amount and leaves the rest of the head to fall into the floor.

Resolve blueprint

A node tree built to be adjusted, not copied blindly

Normalize camera originals into DaVinci Wide Gamut / DaVinci Intermediate with Resolve Colour Management or a documented input CST. Then place black before anything else. In this film the floor is the look: most of the frame sits on it, and every subsequent decision — how far a face is lifted, how bright a lamp may bloom, how much colour survives in the dark — is measured from it.

Keep practical pools, skin and the ambient night colour in separable branches. The ambient teal-green or blue of the darkness and the warm or neon colour of each source are different lights and need independent density. Suggested settings are starting ranges for suitable log or raw footage, never recovered originals, and they assume a monitoring environment dark enough to judge L* 5.

Group Pre-ClipB01

Input transform

Normalize camera and VFX sources correctly before judging a black floor that most of the frame will sit on.

Do this

  • Decode raw from camera metadata, or transform documented log data into DaVinci Wide Gamut / DaVinci Intermediate.
  • Give CG, matte and plate elements their own input transforms before compositing.
  • If project colour management already performs the IDT, omit this node rather than doubling it.

Starting points

  • Keep creative tone mapping out of the input stage; display mapping belongs in B13.
  • For comparison with the reference PNGs only, use Rec.709 gamma 2.4 as an explicit assumption.

Watch

  • A wrong transfer or data/video levels error moves black by more than the whole shadow structure of these frames.
  • Confirm levels before treating a lifted or crushed floor as intentional.
Clip · serialB02

Exposure from the pool

Place exposure from the lit subject inside each pool — a face, a lampshade, a spotlit dress — not from the frame average, which is dominated by black.

Do this

  • Use Offset for exposure, then HDR Global or Log wheels for zone placement.
  • Balance from any credible neutral inside the light, then let the ambient darkness keep its own colour.
  • Never expose to raise the frame average; it is supposed to be low.

Starting points

  • Begin within about ±0.3 stop on well-exposed log material.
  • Read exposure on a face or practical with a waveform zoom, not on the full-frame histogram.

Watch

  • Exposing for the average lifts the floor and turns black into grey.
  • Auto-balance on the ambient removes the cool darkness the frames depend on.
Clip · serialB03

Black floor and toe

Set true black and shape a toe that keeps one or two steps of detail above it — hair, suit, a car roof — without lifting the floor.

Do this

  • Set the floor with Lift or HDR Black; shape the toe separately with Custom Curves or HDR Shadow.
  • Decide per scene what must stay above black (a silhouette edge, a lapel) and keep only that.
  • Keep midtone contrast moderate; the pools supply contrast by themselves.

Starting points

  • Place true black at or just above code zero of the delivery; keep dark costume one or two L* above it.
  • Judge in a dark surround at the target peak; these frames cannot be graded in daylight.

Watch

  • A lifted floor turns the whole film milky and exposes compression.
  • A crushed toe removes the silhouette edges that describe space in the dark.
Clip · serialB04

Soft shoulder for small sources

Hold lamps, spotlights, neon and streetlamps just below clipping with a soft roll, so their small pools keep shape and colour.

Do this

  • Shape a soft shoulder with HDR Highlight/Specular or a custom curve.
  • Leave the sources bright relative to the frame; do not compress them toward the ambient.
  • Use soft clip only as a final guard.

Starting points

  • Start with a shoulder beginning around 80–85% of the scale.
  • Check each source's core keeps its hue instead of going white.

Watch

  • Clipped neon and bulbs turn into white shapes with coloured rims.
  • Over-compressed sources stop reading as light.
Clip · parallelB05

Ambient night colour

Give the darkness its own colour — teal-green in streets and exteriors, blue under a spot — separately from the sources.

Do this

  • Key the dark ambient range with a luminance qualifier and move its hue and density on its own.
  • Keep ambient chroma low; it colours a black, it does not replace it.
  • Exclude practicals, skin and their pools from the ambient key.

Starting points

  • Start with small hue moves in the lower midtones and 10–20% less saturation in the deepest toe.
  • Judge the ambient on the vectorscope at high zoom, since most of its pixels are near centre.

Watch

  • An ambient key that reaches skin turns faces green.
  • A saturated ambient turns black into coloured noise.
Clip · parallelB06

Skin in the dark

Protect skin density and hue inside warm or cold light, and keep faces the brightest broad surface after the practicals.

Do this

  • Key skin with a soft HSL qualifier intersected with face windows; adjust lit-side density only.
  • Keep the shadow side of each face falling into the floor.
  • Under a coloured source (blue spot, neon) keep skin carrying that light rather than neutralizing it.

Starting points

  • Adjust skin by 0.1–0.25 stop and a few degrees of hue.
  • Track windows on moving heads through the whole shot.

Watch

  • A skin key that grabs warm walls or lampshades lifts the whole pool.
  • Neutralizing a face under coloured light makes it look pasted in.
Clip · parallelB07

Practical pools

Hold each practical — table lamp, fringed shade, spotlight, sign — at its own density and colour, with its falloff into black intact.

Do this

  • Key the source and its pool; adjust density and hue separately from the surround.
  • Keep the falloff gradient; it is the only evidence of space in many frames.
  • Let warm practicals stay warm against a cool ambient.

Starting points

  • Start with −0.2 to −0.5 stop inside the practical key and a wide soft edge.
  • Small hue moves only, ±5–10°.

Watch

  • A hard key on a falloff edge produces a visible ring.
  • Brightening a lamp's pool until the room fills with light.
Clip · parallelB08

Neon and narrow-spectrum sources

Keep neon, sodium and coloured spots as saturated sources without letting their colour bleed into the whole frame.

Do this

  • Key each neon colour separately; reduce density at the core, keep hue.
  • Check the red, orange and magenta channels for clipping before adding saturation.
  • Keep spill on nearby faces and cars at a believable fraction of the source.

Starting points

  • Start with 10–20% less saturation in the core and a soft 5–10 px edge.
  • Review defocused neon for chroma noise at full resolution.

Watch

  • Neon cores clipping a single channel and shifting hue.
  • Spill graded up until everything takes the sign's colour.
Clip · serialB09

Local light shaping

Shape attention with soft windows that follow the real light: lift eyes and the lit plane, deepen distracting mid-grey surfaces.

Do this

  • Use soft tracked windows on faces and on the few bright distracting surfaces.
  • Deepen large areas only where the photography already places darkness.

Starting points

  • Start with +0.1 to +0.25 stop on a face window and −0.2 to −0.4 stop on distracting surround.
  • Feather generously; hard windows are obvious against black.

Watch

  • Visible window edges on a black background.
  • Window drift on moving heads.
Clip · parallel · optionalB10

Haze and atmosphere

Keep practical haze visible as a lifted veil around sources without letting it lift the whole floor.

Do this

  • Key haze by luminance in the mid-shadows and adjust density and colour on its own.
  • Preserve the veil around lamps and beams; it is what makes a beam a beam.
  • Starts bypassed; enable only where haze is in shot.

Starting points

  • Start with small density moves, ±0.1–0.2 stop.
  • Match haze colour to the ambient, not to the sources.

Watch

  • Haze key lifting black everywhere.
  • Banding in smooth haze gradients.
Clip · serialB11

Deep-shadow chroma and noise

Keep near-black clean: low chroma, controlled noise, no banding.

Do this

  • Use Lum vs Sat to desaturate the deepest toe.
  • Apply temporal noise reduction before the grade where the source needs it, with care on faces.
  • Add fine grain after grading to break up banding in dark gradients.

Starting points

  • Start with 20–40% saturation reduction below about 5% luma.
  • Temporal NR frames 2–3, low motion estimation thresholds; review faces at 100%.

Watch

  • Coloured noise in black costume and night sky.
  • Waxy faces from aggressive noise reduction.
VFX / Clip · parallelB12

VFX and set-extension integration

Match CG buildings, skies and extensions to the plate's black, haze and grain before the look is applied.

Do this

  • Match element black to the plate's toe and element haze to the plate's atmosphere.
  • Add grain and softness matched to the plate.
  • Check edges in the dark where mattes are hardest to see and easiest to get wrong.

Starting points

  • Start with element black at the plate's floor.
  • Review at speed and frame by frame.

Watch

  • An element whose black is cleaner than the plate's.
  • Mismatched grain between plate and element.
Group Post-ClipB13

Output transform and trim

Render the scene-referred grade to the calibrated delivery while keeping display-specific trims out of the shared look.

Do this

  • Use Resolve Colour Management or a CST from DWG/Intermediate to the required gamut and transfer.
  • Inspect the floor, neon and haze gradients after the transform and after encoding.
  • Create separate SDR, HDR or alternate-display trims rather than changing the scene-referred graph.

Starting points

  • For comparison with the reference PNGs, monitor Rec.709 gamma 2.4 in a dark, controlled surround.
  • Test an 8-bit encode of the deliverable before approval.

Watch

  • A double output transform lifts or crushes the floor.
  • Banding and blocking in near-black once encoded.

8 reconstruction studies

Read the image first; use the scopes to test the read

Each study begins with a literal visible description and tests it against the active-picture measurements of the verified midpoint. Metrics describe the whole frame and cannot isolate a face, a lamp or a sign without the visual inspection recorded in the prose. With deep shadow covering more than half of every frame, frame means are dominated by black; the prose therefore says where the light is as well as what the averages are.

Build the base with every optional node bypassed. Set black and the toe first, then add the named shot operations, check keys and windows at full resolution in a dark room, and inspect moving footage for noise, banding and flicker in the shadows before approving qualifiers, haze or glow.

01 · Study I · The face out of black · shot 4 · 01:01:41:05

A monochrome face rising from nothing

In a black-and-white close-up, a woman with a dark mass of curly hair and one pale curl looks down, lit from above and in front; her bare shoulders and an earring catch the light, and everything around her is black.

Regular-footage feasibility: Conditional Composite/VFX read: none
Black-and-white close-up of a woman with dark curly hair and a single white curl, eyes lowered, bare shoulders and a drop earring lit from above, surrounded by black.
Almost four-fifths of the frame is black; the face is carried by a single overhead light and a few grey steps.

What is remarkable

This is the extreme of the film's floor. 79% of the active picture is deep shadow and the palette's largest cluster is #010101 at 79%; mean L* is 5.3. Yet the face reads completely. The forehead, nose, cheekbones and shoulders sit in a band of mid greys, the eye sockets fall to near black, and the pale curl at upper right is the brightest detail in the frame. The picture is monochrome: mean chroma is 0.5.

The construction is value alone. σL* is 11.0 — wide for a frame this dark — because the face and the black meet at a sharp boundary, and the only transitions are the falloff over the cheeks and the edge of the hair against the void. Nothing else in the frame competes: no background, no fill on the far side of the head.

Where the image comes from

A single overhead key, a black surround with no spill and a performer with makeup that deepens the eyes supply the image; the monochrome is the grade's. Company 3 describes a dedicated black-and-white LUT for these passages, with more midtone contrast than a plain desaturation, and heavy dodging and burning with windows. The grade sets black, places the face's mid greys and chooses a conversion that keeps skin tonality. It cannot create the separation the black surround provides.

Treat the conversion as the palette. Choose channel weights that keep the face's modelling smooth and the curl bright, then set the floor and let everything outside the pool go to true black.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Conditional

Achievable with a clean black surround and a controlled single source; a room with bounce or spill cannot be graded to this black without crushing the face's shadow side.

Minimum capture conditions

  • A black surround with no spill on walls or hair.
  • One key from above and in front, placed for the eye sockets.
  • Low-noise capture so the floor can be set without lifting noise.

Composite and background read

No visible reason to invoke compositing

Hair edges, the earring's specular and the falloff over the shoulders are coherent with a single practical close-up.

Invisible cleanup or background replacement with black cannot be ruled out from a still.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B03

Set the floor to black

Lift/HDR Black so the surround sits at true black; toe shaped so hair keeps one step above it.

Why: Four-fifths of the frame is this floor.

Watch: The hair mass merging into the background.

B06

Place the face's greys

Face window; set the forehead and cheek band to mid grey with smooth falloff.

Why: The face carries the whole frame.

Watch: Harsh contrast in the falloff over the cheeks.

B04

Hold the curl and earring

Soft shoulder so the pale curl and earring highlight keep texture.

Why: They are the only bright details.

Watch: The curl clipping to a white shape.

B11

Keep black clean

Temporal NR on the surround; grain added after.

Why: Noise in a black field is the first thing a viewer sees.

Watch: A waxy face from NR spilling onto skin.

Failure modes to check before approval

  • Noise and blocking in the black surround.
  • The floor lifting to grey in the encode.
  • Loss of the eye sockets' detail after the toe.
  • The curl clipping as the head moves.

02 · Study II · The restaurant · shot 13 · 01:03:23:21

A kiss beside one small lamp

A blonde woman in an orange satin blouse kisses a dark-haired woman across a restaurant table, lit by a small shaded table lamp in front of them; a champagne bottle, glasses and other diners fade into brown darkness.

Regular-footage feasibility: Conditional Composite/VFX read: none
Two women kiss across a dim restaurant table: one blonde with curled hair in an orange satin blouse, the other dark-haired in a dotted dress, beside a small white-shaded table lamp and a green bottle, with other diners blurred in the dark.
One small lamp lights two faces and an orange blouse; the rest of the room is brown-black.

What is remarkable

The lamp is the light source for the whole composition. Its shade, the blonde hair and the orange satin make the warm pool; the palette has near-black #080703 at 60%, dark browns #271506 and #402D15, a burnt orange #3C1503 and one pale warm tone #C19365 at 2%, the shade and the hair. Mean L* is 7.4, 94% of the picture in shadow, and only 2% in highlights.

The faces are in profile and only partly lit: the blonde's cheek and ear take the lamp, the dark-haired woman's profile is rimmed. Background diners sit a few L* above black, softly defocused, so the room has depth without detail. The green bottle in front is almost black but keeps a highlight along its edge.

Where the image comes from

A practical table lamp, probably supported by a soft key from the same direction, warm set dressing and an orange satin costume supply the colour. The grade holds the room brown-black, keeps the lamp shade below clipping, and protects the orange satin from turning red or brown as the toe is printed down.

Expose from the lit cheek and the lampshade. Let the far background fall a step above black and do not lift it to show the room.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Conditional

Achievable when the practical lamp and a motivated key are placed to light the faces within a few stops of the shade; a room lit wider than the lamp's pool cannot be taken down to this without flattening the faces.

Minimum capture conditions

  • A practical lamp in shot, dimmed below clipping.
  • A soft key or bounce motivated by the lamp on the faces.
  • Dark, warm set dressing that falls away without flagging.

Composite and background read

No visible reason to invoke compositing

The lamp's pool, defocus of the diners and occlusion between the two heads are coherent with one practical shot.

Invisible cleanup cannot be ruled out from a still.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B02

Expose from the lit cheek

Offset so the blonde's cheek and the lampshade read correctly; ignore the frame average.

Why: The average is black by design.

Watch: Lifting the room to see the diners.

B07

Hold the lamp's pool

Key the shade and table; lower density slightly and keep the falloff wide.

Why: The lamp is the frame's only source.

Watch: A ring at the pool's edge on the tablecloth.

B06

Protect the orange satin and skin

Separate qualifiers for skin and the blouse before the toe.

Why: Orange satin and skin sit close in hue under a warm lamp.

Watch: The blouse turning red-brown in its folds.

B03

Brown-black room

Toe so the background diners keep one step above black.

Why: Depth without detail.

Watch: Background noise becoming visible.

Failure modes to check before approval

  • Lampshade clipping and losing its shape.
  • Orange satin and skin converging.
  • The dark-haired woman's profile disappearing into the background.
  • Coloured noise in the brown shadows.

03 · Study III · The fringed lamp · shot 98 · 01:06:56:03

A man turning from a rose-coloured shade

An older man in a dark pinstriped suit looks back over his shoulder; behind him at left a fringed lampshade glows rose-orange, a bulb flares at the top corner, and a woman and another man sit blurred in the gloom.

Regular-footage feasibility: Conditional Composite/VFX read: none
An older grey-haired man in a dark pinstriped suit looks back over his shoulder in a dark room, with a glowing rose-orange fringed lampshade and a bright bulb out of focus at left and two blurred people behind him.
The shade glows, the face is lit from the front, and the suit falls into the floor.

What is remarkable

Mean L* 5.3 and 73% deep shadow: a frame almost as dark as the monochrome face, but with colour in it. The fringed shade is the warm event, #73300E at 6% of the palette with a few more-chromatic pixels (high-chroma share 4.8%), and the man's face, lit from the front-left, is the only broad surface above the mid-shadows. The pinstripe survives as a faint texture on a near-black suit.

The colour temperature of the room is ambiguous by design: the shade is rose-orange, the bulb above it nearly white, and the background figures a brown-grey. The face reads warm-neutral rather than taking the shade's rose, which says the key on him is not the lamp.

Where the image comes from

A practical fringed shade in frame and a separate soft key on the face supply the image; the shade is decoration and motivation, not the light on him. The grade keeps those two apart: the shade's rose held as a source, the face kept neutral-warm, the suit falling into the floor with its stripe just visible.

Separate the shade from the face with its own key. Hold the bulb's flare below white. Resist lifting the background figures; their blur and low level are the room.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Conditional

Achievable with a practical shade dimmed to its colour and an independent key on the face. A shade used as the only light would put its colour on the face, and the grade cannot remove that without flattening skin.

Minimum capture conditions

  • A dimmed practical recorded below clipping with its fringe texture.
  • A separate soft key on the face.
  • Dark wardrobe with a fine pattern that holds above the floor.

Composite and background read

No visible reason to invoke compositing

Defocus, the bulb's flare and the lamp's position relative to the actors are coherent with one practical setup.

Invisible cleanup cannot be ruled out from a still.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B07

Hold the shade as a source

Key the shade; keep its rose-orange and lower core density slightly.

Why: The shade is the only colour in the frame.

Watch: Rose turning red or orange as saturation changes.

B06

Keep the face neutral-warm

Face key excluding the shade's hue; lift the lit plane 0.1–0.2 stop.

Why: The key on the face is not the lamp.

Watch: Skin taking the shade's rose.

B04

Contain the bulb

Soft shoulder for the flare at top left.

Why: An unconstrained flare competes with the face.

Watch: A white clipped disc with a coloured rim.

B03

Suit into the floor

Toe so the pinstripe stays faintly visible.

Why: The suit is the largest dark shape next to the face.

Watch: Stripes disappearing or the suit lifting to grey.

Failure modes to check before approval

  • Face taking the lampshade's rose.
  • Bulb flare clipping.
  • The pinstripe vanishing after noise reduction.
  • Background figures lifting into view.

04 · Study IV · The club · shot 645 · 01:37:05:11

Dancers in a single blue spotlight

In a black club, a man in a top hat and tails and a woman in a red-pink feathered dress dance under one blue spotlight whose cone falls from a point high in the frame and pools on the floor around them.

Regular-footage feasibility: Conditional Composite/VFX read: low
A man in a top hat and tails and a woman in a feathered red dress dance in a cone of blue light falling from a single spotlight above, in an otherwise black nightclub with faint figures at the edges.
One blue cone carries the frame; the red dress is lit blue and reads magenta-pink.

What is remarkable

This is the darkest frame here and the most saturated: mean L* 4.6 with 77% deep shadow, and mean saturation 0.98 — the few pixels that are lit are almost entirely blue. The warm/cool balance is 0.0: there is no warm pixel above the chroma threshold in the frame. The palette runs from black #000105 through navy #020A23 and #001843 to a mid-blue #12357A and a bright #2C7EC4 at 2%, the cone's core.

The dress is the interesting case. Under a pure blue source a red dress should go dark, and much of it does, but its lit feathers read magenta-pink; the blue light and the dress's own colour mix at the surface. The cone is visible because of haze, its edge hard at the top and softening as it widens.

Where the image comes from

A single blue spotlight with haze, a black stage and costume supply the construction; this is lighting, not grading. The grade holds the cone's gradient and the pool on the floor without clipping blue, keeps black under it truly black, and keeps the dress's magenta rather than neutralizing the scene.

Grade the blue as a source. Blue is the channel most likely to clip and to band in an 8-bit delivery, so the shoulder and the haze gradient are the main work.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Conditional

Achievable with a real blue source and haze; the cone and pool cannot be convincingly added in the grade, and a blue wash applied to a white-lit scene would not produce the dress's magenta.

Minimum capture conditions

  • A saturated blue source with its core below clipping in the blue channel.
  • Controlled haze to reveal the cone.
  • A stage or club dark enough that the cone is the only light.

Composite and background read

Practical lighting is the plausible read

The cone's origin, its widening through haze and the pool on the floor are coherent with a real follow-spot. Beams are sometimes enhanced in compositing, but nothing here requires it.

Beam or haze enhancement cannot be ruled out from a still.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B08

Hold the blue core

Key the cone's core; reduce density 0.2–0.4 stop and 10–20% saturation.

Why: Blue clips first and bands worst.

Watch: A cyan-white hotspot at the source.

B10

Keep the cone's gradient

Haze key in the lower mids; small density moves only.

Why: The cone is visible only through haze.

Watch: Banding across the cone in the encode.

B06

Leave the dress magenta

No neutralizing of the dancers; small lift on the lit plane of faces if needed.

Why: The mixture of blue light and red dress is the frame's colour event.

Watch: Pulling the dress back toward red.

B11

Clean black around the cone

Desaturate below 5% luma; temporal NR on the surround.

Why: Blue light makes coloured noise in black very visible.

Watch: Blue speckle in the black stage.

Failure modes to check before approval

  • Blue channel clipping in the spotlight's core.
  • Banding across the haze cone.
  • Blue noise in the surrounding black.
  • The dancers' faces lost inside the cone.

05 · Study V · The masked man · shot 657 · 01:37:37:03

A masked face against defocused neon

A man in a black hat and a black knitted mask over his nose and mouth stands at right, one eye lit; behind him at left, out of focus, glow shapes of magenta, violet, green and amber light.

Regular-footage feasibility: High Composite/VFX read: none
Close on a man in a black hat and black knitted face mask, only his eye and part of his brow lit, against blurred magenta, violet, green and amber lights at the left of the frame.
The man is almost all black; the colour lives in the defocused lights behind him.

What is remarkable

The frame is split: the right half is the man, nearly all black, and the left is colour without objects — defocused magenta #7B366A and #481732, a green band, an amber blob #BC7323 below. Mean L* 7.3 with 70% deep shadow and 8.3% high-chroma pixels, all of them in those lights. A green cast touches the brow and the edge of the eye.

The face is reduced to a single eye and a sliver of brow; the mask, hat and coat merge into one black shape defined only by its edge against the lights. The eye is the subject, and it reads because it is the only lit skin in the frame.

Where the image comes from

Coloured practicals or neon at a distance, a long lens at wide aperture, and a black-costumed performer supply the image. The grade holds each neon colour in its own range, keeps the black shape of the figure clean against it, and puts a small lift on the eye.

Keep the colours separated: magenta, green and amber must not converge into a brown blur. Hold the black silhouette at the floor; its edge against the lights is the drawing.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

High

Straightforward with neon or coloured practicals behind the subject and a fast lens; the defocus and colour come from the photography and the grade only keeps them apart.

Minimum capture conditions

  • Coloured sources several metres behind the subject.
  • A long lens at wide aperture.
  • A small eye light that does not spill onto the mask.

Composite and background read

No visible reason to invoke compositing

The bokeh shapes, their overlap with the hat's edge and the green spill on the brow are coherent with practical capture.

Invisible cleanup cannot be ruled out from a still.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B08

Separate the neon colours

Individual keys on magenta, green and amber; hold hue, lower density at the cores.

Why: Defocused colour turns muddy when compressed together.

Watch: Magenta and green blending to grey at their overlap.

B03

Black silhouette

Floor and toe so the hat, mask and coat merge into one clean black shape.

Why: The edge against the lights is the drawing.

Watch: Lifted blacks revealing the knit texture as noise.

B09

Lift the eye

Small tracked window, +0.1–0.2 stop.

Why: The eye is the only subject.

Watch: Window visible on the brow.

B11

Clean deep shadow

Desaturate below 5% luma.

Why: Coloured light near black produces chroma noise.

Watch: Magenta speckle on the mask.

Failure modes to check before approval

  • Neon colours converging into a brown blur.
  • Chroma noise on the black mask.
  • The eye light drifting as the head moves.
  • Amber blob clipping in its core.

06 · Study VI · The street · shot 1933 · 02:33:12:22

Two figures walking into haze

At night, a woman with a pale mass of curly hair and a man in a jacket walk down a street lined with parked 1930s cars, backlit by streetlamps through haze; a car's headlights glow at the far end.

Regular-footage feasibility: Conditional Composite/VFX read: low
Two figures, a woman with pale curly hair and a man in a jacket, walk down a hazy night street between parked vintage cars, backlit by glowing streetlamps with another car in the fog at the far end.
Haze and backlight turn the street into layers; the figures and cars are silhouettes against a green-grey veil.

What is remarkable

Depth here comes from haze. Near cars are black shapes with rim highlights, the figures are silhouettes, and the far car and the street beyond fade into a green-grey veil — #162621 at 22% and #031513 at 13% of the palette, with the frame's mean a* negative (−2.9). The lamps are warm, #83745C and above, and the only warm points. Mean L* 9.7 with 56% deep shadow: brighter than most of this report because the haze lifts the background.

The woman's pale hair catches the backlight and is the brightest broad shape on the figures; the man's face is barely readable. The wet road carries a soft sheen from the lamps.

Where the image comes from

A dressed street with period cars, practical streetlamps, haze and a backlight at the end of the street supply the construction. The grade keeps the haze a veil rather than a lift of the whole floor, holds the ambient green-grey, and keeps the lamps warm and below clipping.

Separate the haze layer from the near blacks. The near cars must stay nearly black; only the air behind them is lifted.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Conditional

Conditional on real haze and backlight in the street; a grade can shape a veil but not invent the layered separation that atmosphere gives near and far objects.

Minimum capture conditions

  • Practical haze with a controlled density along the street.
  • A backlight at depth and warm practical streetlamps.
  • Dark foreground objects to hold the floor.

Composite and background read

A dressed practical street is the plausible read

The haze's interaction with the lamps, the rim light on the cars and the figures' contact with the road are coherent with one plate. Street depth is sometimes extended digitally in period work, but the far end here is fog.

A still cannot rule out set extension beyond the haze.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B10

Haze as a veil

Luminance key on the mid-shadows behind the figures; hold its density and green-grey colour.

Why: The haze is the frame's depth.

Watch: The haze key lifting the near cars.

B05

Green-grey ambient

Small hue move on the dark ambient toward the measured green-grey; low chroma.

Why: The ambient colour of the street is cool and separate from the lamps.

Watch: Green spilling onto the woman's hair.

B07

Warm lamps below white

Key the lamps; keep them warm with a soft core.

Why: They are the only warm light.

Watch: Clipped lamps with halos in the haze.

B03

Near objects on the floor

Toe so the foreground cars read as black shapes with rims.

Why: Layering depends on near black against the veil.

Watch: Near cars lifting into the haze.

Failure modes to check before approval

  • Haze banding in the encode.
  • Near cars lifting and flattening the layers.
  • Lamps clipping into white blobs.
  • Flicker in haze density across the shot.

07 · Study VII · The gas station · shot 2004 · 02:38:51:15

Orange neon at dusk

At dusk, police cars with small red lights gather in front of a white clapboard gas station under an orange OLEX neon sign; a man stands in a lit telephone booth at left, globe pumps glow at right, and dark trees fill the sky.

Regular-footage feasibility: Conditional Composite/VFX read: medium
A roadside gas station at dusk with an orange OLEX neon sign, several dark 1930s police cars with red roof lights, uniformed officers, a man inside a lit telephone booth at left, and glowing globe fuel pumps at right in front of dark trees.
A teal dusk with a single warm family: orange neon, red roof lights and the lit booth.

What is remarkable

This is a blue-hour exterior carried almost entirely by the ambient: 60% of the frame is deep shadow and the palette is dominated by dark teal-blues — #03080A, #051C28 at 22%, #244C56 — with the warm family, #604E36 and #B8774A, making up 10%. Mean L* 9.8, warm/cool balance 0.12. The neon lettering, the orange tube along the roofline and the second neon at the right edge are the brightest warm objects; the cars are blue-black shapes with chrome.

The white clapboard of the station reads pale teal: it is lit by the dusk ambient, not by its own sign. The telephone booth at left has its own interior light, a slightly green-yellow pool with a man inside, a third colour temperature in the frame.

Where the image comes from

A built or dressed gas station, practical neon, working car lights and a dusk sky or dusk-for-night lighting supply the image. The grade holds the ambient teal-blue at low density, keeps the neon orange rather than red, and keeps the booth's green-yellow distinct. It controls the time of day: the scene can be pushed from dusk to night by density alone.

Place the neon first, then the ambient. The station's white siding is the frame's neutral reference and should read as white under blue light, not as blue paint.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Conditional

Conditional on shooting at dusk or lighting for it with practical neon; a daylight exterior graded down would lose the neon's glow and the ambient's balance.

Minimum capture conditions

  • Real neon and practicals recorded below clipping.
  • A dusk ambient or a large soft cool source across the set.
  • Dark trees or sky behind to hold the upper frame down.

Composite and background read

A dressed location with possible extension

The station, cars and actors are coherent with a practical set. The tree line and sky above are dark and featureless, which is where extension or darkening is often applied, and the right-hand neon at the frame edge could be set dressing or digital. Multiple period cues and the controlled ambient make some environment work plausible.

Nothing in the still is decisive; plates or production records would be needed to confirm any extension.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B08

Orange neon, not red

Key the neon; lower core density and keep hue toward orange.

Why: The neon is the frame's warm anchor.

Watch: Red channel clipping and shifting the neon toward red.

B05

Teal dusk ambient

Ambient key on the lower mids; hold teal-blue at low chroma.

Why: The ambient defines the time of day.

Watch: Teal spilling into the booth's pool.

B07

Keep the booth separate

Its own key; retain its green-yellow.

Why: A third light temperature gives the frame depth.

Watch: Neutralizing the booth into the ambient.

B12

Match the upper frame

If trees or sky are extended, match their black and grain to the plate.

Why: The upper frame is where extensions show.

Watch: A cleaner black above the treeline than below.

Failure modes to check before approval

  • Neon clipping in the red channel.
  • Station siding turning blue rather than white under blue light.
  • Chrome highlights clipping as cars move.
  • Grain mismatch between plate and any extended sky.

08 · Study VIII · The tower · shot 2327 · 02:56:45:10

A lit tower that was almost certainly built

At night, a rounded building rises to a glass-walled tower glowing cyan, a spark of electricity at its top; around it dark skyscrapers with a few lit windows, an ONYX sign in orange-red and dark foliage at right.

Regular-footage feasibility: Low — useful reference Composite/VFX read: high
Low-angle night view of a rounded dark building topped by a glowing cyan glass tower with a small electrical spark above, surrounded by dark skyscrapers with scattered lit windows, an orange ONYX sign at left and trees at right.
A teal-black city in which the tower glows cyan and one small sign is the only warm light.

What is remarkable

Mean L* 5.2 and 71% deep shadow, but with a saturation of 0.94: what little is lit is strongly coloured, and nearly all of it is teal — #021818 at 23%, #032F37 and #266266. The warm/cool balance is 0.01. The ONYX sign and a few window lights are the only warm pixels. The tower's glass reads pale cyan with its mullions visible, and a small electrical spark at its top is the brightest point in the frame.

The surrounding buildings are dark silhouettes with scattered lit windows, the whole city receding into a teal haze. The image reads as an establishing view and has very little of the texture or noise of the handheld street frames in this report.

Where the image comes from

The construction is very probably environment work: a tower, city and spark built or extended digitally, perhaps around a practical foreground building. The grade unifies the teal ambient, holds the tower's glow below clipping and keeps the ONYX sign warm; it cannot supply the architecture.

Integrate in scene-referred space before the look, then grade the whole as one night: black at the floor, a teal veil above it, the tower as the source.

Waveform

RGB distribution

Vectorscope

Can regular footage get there?

Low — useful reference

Settling on a useful reference rather than a generally reproducible candidate. The grade is simple; the environment, glow and spark require visual effects that no grade can supply from ordinary footage.

Minimum capture conditions

  • A modelled or photographed tower and city elements.
  • A glow and spark element with matching atmosphere.
  • Compositing before the grade with matched grain and black.

Composite and background read

Environment work is strongly plausible

An electrical spark at the tower's top, a glowing glass cylinder, a skyline of dark towers with evenly scattered lit windows and very low image noise for a frame this dark together point to a constructed establishing shot.

This is a reading of construction, not a production credit; which parts were photographed cannot be determined from a still.

Shot-specific Resolve pass

Use the global tree, then make these decisions

B12

Integrate the environment

Match element black, haze and grain before the look.

Why: The frame is mostly elements.

Watch: Cleaner blacks in elements than in any plate.

B05

One teal night

Ambient key across the whole frame; teal at low chroma.

Why: A single ambient unifies the city.

Watch: Different teal families in different buildings.

B04

Hold the tower and spark

Soft shoulder for the tower glow and the spark.

Why: They are the source and the event.

Watch: The glass clipping to a flat white shape.

B08

Keep ONYX warm

Separate key on the sign; hold its orange-red.

Why: The only warm light in the frame.

Watch: The sign going pink under the teal ambient.

Failure modes to check before approval

  • Tower glow clipping and losing its mullions.
  • Grain mismatch between elements.
  • Banding in the teal haze around the tower.
  • The spark's brightness changing frame to frame.

Measurement note. Measurements use the active 1920×800 image only. L* and C* follow the provisional gamma-2.4 interpretation; percentages describe this verified midpoint, not a duration or an object mask. In frames this dark, a difference of a few L* in the shadows is a handful of 8-bit code values.

Reconstruction note. Suggested Resolve values are broad first-pass ranges for correctly normalized log/raw sources. Reconstruct value and colour relationships — where black is, what sits one step above it, how small each pool is — rather than knob positions, and make separate output trims for the actual delivery.