Raking Light

An original design for hard tenebrism: one high source raking across the subject, flesh that runs hot where the light lands, and a ground that keeps no colour at all. Derivation in the notes.

Painterly reference: Caravaggio, measured from eleven public-domain reproductions against eleven other painters. Named for the lighting rather than the painter, which is this bank's standing rule.

Provenance. Built 2026-09-17 from eleven Caravaggio oils, public domain, photographed cropped to the canvas and verified by eye. Measured in CIELAB against eleven other painters gathered the same way.

The most extreme signature in the control group

Two numbers in that column do all the work.

The lit passages are hot. Highlight a* +10.06 with b* +24.83 is the warmest lit tone of any painter measured, by a clear margin over Rembrandt at +6.67 and +19.40. Where light lands on flesh in these pictures, it runs red and orange rather than merely bright.

The unlit passages have no colour in them. Shadow b* +0.49 is effectively zero. The dark is not a dark brown or a warm umber. It is an absence.

The gap between the two, −24.34, is the largest in the twelve-painter group and a third larger than Rembrandt's. A second estimator using fixed absolute bands and a median returns −20.42, and the per-image sign test is unanimous: zero of eleven paintings fall within three units of a neutral split. Nothing here is borderline.

The palette is correspondingly narrow. Red-orange 45.9% and yellow 52.9% account for 98.8% of coloured pixels. Green measures 0.2%, cyan-blue 0.1% and violet 0.0%.

Why Classic Chrome, which is the surprise

The obvious base for a Old Master look is Reala Ace, and this bank already used it that way for Dark Ground. The measurement points somewhere else.

• A hard shadow end is the first requirement, against a black point of 5.3 and a median of 15.5. Fujifilm publishes Classic Chrome as hard at the shadow end and second only to Velvia in overall contrast, and this bank's Dynamic Range & Tone Curve, the contrast controls tone table notes it needs a Shadow lift merely to reach normal shadow detail. Declining that lift is most of this recipe.
• Cool shadows, which are normally a thing to correct, are correct here. Fujifilm's own comparison against Provia is "suppressed magenta and cool shadows", and the corpus shadow measures b* +0.49 against a highlight of +24.83. A base that keeps its darks free of warmth is reproducing the reference rather than fighting it.
• Suppressed magenta costs nothing, because magenta is 0.0% of this palette.
• The soft highlight end holds a white point of 66.1, which is well short of clipping. Lit flesh should glow rather than blow.

How it differs from Dark Ground, which is the obvious objection

Dark Ground is this bank's existing Old Master portrait recipe, and three painters in the control group (Rembrandt, Titian, El Greco) measure close enough to it that they were not built. Caravaggio was, and the reason is a specific difference rather than a preference.

Dark Ground's brief is skin that is "luminous without being warm" on a warm dark ground. The Caravaggio corpus measures the reverse on both counts: highlights at a* +10.06 and b* +24.83 are conspicuously warm, and shadows at b* +0.49 are not warm at all. One recipe puts the warmth in the ground and keeps it out of the skin. This one puts it in the skin and keeps it out of the ground.

The settings follow that split. Dark Ground runs Reala Ace at DR200 with Highlight 0 and Shadow −1, softening the bottom. This runs Classic Chrome at DR100 with Shadow +1, hardening it, and adds three notches of Red where Dark Ground adds one.

The rest

• DR100, and it is doing the same job here as in Pink Chrome. Every Dynamic Range mode underexposes and then lifts everything below the highlights. The black point is 5.3 and the median is 15.5, so a DR mode would lift almost the whole picture. There is also nothing to protect at the top, since the white point is 66.1. The bonus is base ISO 125.
• Shadow +1 on a base that is already hard at the shadow end. This is the most aggressive shadow setting in the creative folder and it is aimed at a median of 15.5, which is darker than any other target this bank has built for. If it crushes texture you wanted, the fix is to move the light, not the curve.
• Exposure −1. A meter presented with a frame that is four fifths dark will try to make it middle grey, which is the single most reliable way to lose this look. Go to −1 1/3 when the ground dominates, and check that the histogram is piled against the left with a thin tail to the right, because that is what the corpus is.
• White Balance +3 Red and −2 Blue, the strongest Red shift in this folder. It is aimed at a highlight a* of +10.06, and it works because the shadows have almost no chroma for it to act on. A warm shift moves the lit flesh a long way and leaves an empty dark where it is, which is the split the look depends on.
• Color +2 with Color Chrome Effect Strong. Classic Chrome is a low-saturation base and the target is 14.0, so Color lifts the lit passages toward it. Colour Chrome Effect then puts gradation inside them, which is what stops a hot cheek or a red drape becoming a flat shape. Neither control has much to act on in the dark.
• Color Chrome FX Blue Off. Cyan-blue measures 0.1% and violet 0.0%. There is nothing to deepen, and the one thing FX Blue could reach is a cool rim on a shadow that should stay empty.
• Sharpness 0 and Clarity −1. Skin is the subject and the reference is oil paint at viewing distance. Clarity is also the setting most likely to dig contrast back into a shadow that this recipe has worked to empty.

Variants

Movie mode

Ports reasonably. Movie mode drops Grain, Color Chrome Effect and Clarity. Grain and Clarity are near-neutral here, so the real loss is Color Chrome Effect Strong, which means the lit passages will read flatter and the hot skin more like a single block of colour. Classic Chrome, the white balance, DR100 and the tone curve all carry, and those are what produce the split. Light for it rather than grading for it and the look survives. See Video / Movie Mode, settings, differences, and photo→video translation.

Validation tier: Creative design, field-test pending. Grounded in Fujifilm's published design statements for Classic Chrome and for each base rejected above, in this bank's Dynamic Range & Tone Curve, the contrast controls tone-curve table and Color, Color Chrome Effect & FX Blue, the saturation/richness controls colour control notes, and in a CIELAB measurement of eleven verified reproductions against an eleven-painter control group, with two estimators and an unanimous per-image sign test.

No frames have been shot with it.

Test priority: in order

1. One hard source high and to one side, a face, no fill at all. The whole recipe in one frame. Measure the lit cheek and the background separately: the targets are roughly a* +10 with b* +25 on the flesh, and b* near zero in the dark.
2. The gap between those two b* values. The target is about −24, which is the largest split this bank has ever built toward. If it comes back nearer −12, the fill in your room is doing something you did not intend, and the answer is a flag rather than a setting.
3. The histogram. Four fifths of it should be piled left of the middle with a thin tail to the right. A centred histogram means the meter won and the frame is not this look.
4. Skin at +3 Red. The most likely thing to have overshot, and the variants table carries the retreat. Pale skin reaches orange sooner than the reference does, which was measured off Mediterranean flesh under a warm source.
5. Any dark area at 100 per cent. The Shadow +1 check. The ground should be empty of colour rather than full of noise, and High ISO NR −4 plus an underexposed frame is the combination that can put grain where the recipe wants nothing at all.

What would count as the recipe being wrong rather than needing a tweak: if an empty shadow comes back noisy rather than clean at base ISO, then a sensor's noise floor is doing what varnish and glaze did for a painter, and the look needs a lifted black with the colour drained out of it instead, which is a different tone curve and a different argument.

Pro tip: This is a lighting recipe wearing a settings table. The measured split of −24.34 between lit and unlit is produced by a single hard source with no fill, and no combination of camera settings invents it. Move the lamp first. If you have a second light on, you are not shooting this.

Pro tip: The dark is supposed to be empty, not brown. The most common way to miss Caravaggio is to make the shadows warm, because that is what Rembrandt does and what most people picture. The corpus measures shadow b* +0.49 against highlight b* +24.83. Warmth belongs on the skin and nowhere else.

Pro tip: Meter down, hard, and trust the histogram rather than the screen. A median L* of 15.5 is darker than anything else in this bank, and every automatic system in the camera is built to prevent it.

Pro tip: Classic Chrome for a Old Master look is counterintuitive and it is what the numbers say. Its cool shadows are usually a thing to correct. Here they are the reference. Its suppressed magenta usually costs something. Here the palette has 0.0% magenta so it costs nothing. A base is only wrong relative to a subject.

Shoot JPEG. HEIF silently drops Clarity. See Grain & Detail. Grain Effect, Sharpness, Clarity, High ISO NR.

Creative design (field-test pending). See the notes and the Knowledge base.

Part 0: method

Third recipe from the painting seam, built with the same instrument and the same control group as Open Air and Chrome Yellow: twelve painters measured in CIELAB on one day, only differences between them treated as findings.

That design exists because a single painter's reproductions cannot be read alone. Museum photography is colour-managed but not identical between institutions, and four centuries of varnish sits between an old master's palette and the file you download. Varnish yellowing matters more for Caravaggio than for anyone else in the group, and Part 5 says what that does to these numbers.

This recipe was built after the corpus problems that forced corrections to the first two were found and fixed, so it is the first in the seam measured on a clean pipeline from the start. The full account of those problems is in the notes.

Part 1: the corpus

Eleven Caravaggio oils, public domain, photographed cropped to the canvas, every one verified by eye. One duplicate was removed by perceptual hash.

Eleven is a small corpus and it is stated rather than glossed. Caravaggio's surviving autograph output is itself small, and the subset with open-access high-resolution reproductions is smaller. What makes eleven workable here is the size of the effect: the finding below is unanimous across every individual painting, which a marginal result would not be.

The images were fetched into a scratch workspace and are not in this collection.

Part 2: the measurement

Finding 1: the widest split in the group, and it is unanimous

The last column asks how much yellowness the dark parts of a picture carry relative to the bright parts. Caravaggio measures −24.34, the largest of twelve painters, a third larger than Rembrandt and eleven times Monet's.

Two properties combine to produce it, and both are extremes in their own right:

• Highlight a* +10.06 with b* +24.83. The warmest lit tone in the group, well clear of Rembrandt at +6.67 and +19.40. Light landing on flesh runs red-orange rather than simply bright.
• Shadow b* +0.49. Effectively zero. The dark is not warm brown or umber. It holds no colour at all.

The finding survives a change of instrument. A second estimator using fixed absolute luminance bands and a median in place of per-image quintiles returns −20.42, preserving both the magnitude and the ranking.

It also survives the per-image sign test, which is where eleven paintings becomes enough. Zero of eleven Caravaggios fall within three units of a neutral split, against nine of sixteen Monets. There is no borderline case in this corpus.

Finding 2: two hue families and nothing else

Red-orange and yellow together are 98.8%. This is the narrowest palette in the twelve-painter group, narrower even than Rembrandt's 95.1%, and it settles Color Chrome FX Blue without further argument.

Finding 3: the darkness

A median L* of 15.5 means half of a typical frame sits below fifteen per cent lightness. For comparison, Pink Chrome, which this bank described as having the deepest black in its reference work, has a black point of 7.6 but a median far higher. Caravaggio's black point of 5.3 is comparable. What is different is how much of the picture lives down there.

That distinction matters for the settings. A deep black point is reached with exposure. A dark median is reached by lighting a small part of the frame and leaving the rest, which is why the notes says plainly that this is a lighting recipe wearing a settings table.

Part 3: the base, and why it is not the obvious one

The obvious base for a Old Master look is Reala Ace, which this bank used for Dark Ground. Four requirements come out of Part 2 and they point elsewhere.

1. A hard shadow end, for a black of 5.3 and a median of 15.5.
2. Shadows that stay free of warmth, for a shadow b* of +0.49 against a highlight of +24.83.
3. A highlight end that rolls rather than clips, for a white point of 66.1.
4. Nothing spent on hues the palette does not contain.

Classic Chrome meets all four, and the second is the interesting one. Fujifilm's published comparison against Provia is "suppressed magenta and cool shadows". Cool shadows are normally a property to correct, and several recipes in this bank spend a white-balance shift doing exactly that. Here they are the reference. A base whose darks decline to warm up is reproducing Caravaggio rather than fighting him.

This bank's Dynamic Range & Tone Curve, the contrast controls tone table adds the first requirement: Classic Chrome is hard at the shadow end, second only to Velvia overall, and needs a Shadow lift merely to reach normal shadow detail. Declining that lift, and then adding Shadow +1 on top, is most of this recipe.

Requirement four is free. Suppressed magenta costs nothing when magenta is 0.0% of the palette.

Part 4: the separation between this and Dark Ground

Three painters in the control group measure close enough to Dark Ground that they were not built: Rembrandt, Titian and El Greco. Caravaggio was built, and the case rests on a measured difference rather than on a preference for variety.

Dark Ground's stated brief is skin "luminous without being warm" over a warm dark ground. The Caravaggio corpus measures the reverse on both halves:

One recipe puts the warmth in the ground and keeps it out of the skin. This one does the opposite. That is a genuine inversion rather than a matter of degree, and it is why the two coexist.

The settings follow. Dark Ground runs Reala Ace at DR200 with Highlight 0 and Shadow −1, softening the bottom, and one notch of Red. This runs Classic Chrome at DR100 with Shadow +1, hardening it, and three notches of Red.

Part 5: honest limits

• No frames have been shot with it.
• Eleven paintings is a small corpus. It carries this finding because the effect is unanimous across every individual work, and it would not support a fine distinction.
• Varnish cuts against this recipe more than any other in the seam. Four hundred years of darkened varnish both lowers the measured lightness and warms the measured highlights, so some part of both headline figures belongs to the coating rather than the paint. The direction is knowable even if the size is not: a cleaned Caravaggio would measure lighter and slightly less warm at the top. The recipe is therefore aimed at the pictures as they hang, which is how anyone forming an impression of Caravaggio has seen them, and that is the honest framing rather than a claim about what left the studio.
• The white balance was measured off Mediterranean flesh under a warm source. Three notches of Red will reach orange sooner on pale skin, which the variants table addresses.
• A reference tells you what a look is and never whether a recipe reproduces it.
• The split this recipe targets is produced by lighting, not by settings. No camera configuration invents a −24 b* gap in a room with fill in it.

Part 6: where it sits in the bank

Third painting-derived recipe, after Open Air and Chrome Yellow, and the three are as far apart as the seam allows: a black point of 34.9 against 22.1 against 5.3, and a shadow-to-highlight split of −2.24 against −14.50 against −24.34. Monet and Caravaggio sit at the two ends of the twelve-painter range on the measure that matters most, which is a reasonable argument that the seam is worth continuing.

Third creative original on Classic Chrome, after Austurbane and Clerestory, and the first to use the base for its cool shadows as a feature rather than as something to correct.

It is the darkest recipe in the bank by median, and the only one running Shadow positive on a base that is already hard at the shadow end.

Last updated: 2026-09-17