Open Air

An original design for the Impressionist daylight look: no black anywhere in the frame, shadows that carry colour instead of losing it, violet in the darks and a general high-key lift. Derivation in the notes.

Painterly reference: Claude Monet and the plein-air Impressionists, measured from sixteen public-domain reproductions against eleven other painters. The recipe is named for the working method rather than the painter, which is this bank's standing rule and which matters more than usual here: this is a measured property of a movement, not a claim to reproduce one man's hand.

Provenance. Built 2026-09-17 and re-measured the same day against a corrected corpus. The first pass took reproductions on filename alone and admitted four canvases photographed in their gilt frames, which inflated chroma and the yellow share. Every figure below comes from the second pass: sixteen Monet oils cropped to the canvas, verified by eye, measured against eleven other painters. What changed and what survived is in the notes.

The measurement, and why the control group matters

Measuring one painter tells you very little, because every reproduction carries the museum's lighting and the varnish's age. Measuring twelve and comparing them tells you what is specific to one. The nine best-sampled are below.

Three of those columns carry the recipe.

The black point. Monet measures 34.9 where Rembrandt is 4.7 and Caravaggio 5.3. Nothing in these frames comes near black, and the gap to the old masters is a different register of picture rather than a matter of degree.

The shadow-to-highlight warmth difference. This column asks whether the dark parts of a picture carry the same yellowness as the bright parts. The other eleven painters run from −9.3 to −24.3, meaning their shadows carry far less colour temperature than their lights: dark areas are where colour drains out. Monet measures −2.24, four times closer to zero than the next painter and eleven times closer than Caravaggio. A second estimator using fixed absolute bands and a median rather than per-image quintiles returns −3.04, so the ordering survives a change of instrument.

That is the numeric form of the thing Impressionism is actually known for, painting shadows as coloured light rather than as an absence of light, and it turns out to be measurable rather than merely assertable. It is a difference of degree rather than an absolute, and the notes records that an earlier pass overstated it.

The violet. Blue-violet is 26.9% of coloured pixels, twice the next painter and effectively absent in four of the eleven. It is the second-largest hue family in the corpus, behind yellow-green at 29.6% and ahead of yellow at 26.7%. Combined with the warmth result, these shadows are violet and warm at the same time rather than neutral-dark.

Why Astia

Four requirements fall out, and one simulation meets all four.

• Soft at the shadow end, badly needed. A black point of 34.9 is by a wide margin the most lifted target this bank has ever built for. Fujifilm publishes Astia as soft at both ends, so the base starts where the recipe needs to end up.
• Soft at the highlight end too. A white point of 77.8 is well short of paper, so the top has to roll rather than clip.
• Saturation maintained. Fujifilm's phrasing for Astia is soft at both ends "while saturation is maintained", and 11.9 is gentle rather than drained. Astia is the base that holds a low chroma without flattening it.
• No per-hue opinion. The palette runs yellow, yellow-green and violet together, and a base that reshapes hues toward memory colour would rearrange the relationship the measurement found.

The setting that is a zero on purpose

White Balance sits at 0 Red, and that is the unusual decision in this table.

Almost every creative recipe in this bank moves the Red axis, because almost every look has a cast. This one measures shadow a* at −1.14 against highlight a* at −0.99, which is the same very slight cyan lean at both ends rather than any red at all. A Red shift would invent a cast the corpus does not have, and would work against the violet that Part 3 of the notes identifies as the signature.

The single notch away from Blue supplies the mild uniform warmth the numbers do show, shadow b* +3.21 against highlight b* +5.45, and that is all it does.

The rest

• DR400 with Shadow −2. This is the most shadow-lift in the bank and it is aimed at a measured 34.9, which is higher than the first pass suggested and makes the pairing less optional than it looked. DR modes underexpose and then lift everything below the highlights, which is normally a cost paid for highlight headroom. Here the lift is the point and the headroom is a bonus. The floor lands at ISO 500, which is this sensor's conversion-gain step and therefore the cheapest place in the range to shoot DR400.
• Highlight −1.5 rather than −2. This bank's tone table warns against stacking highlight softening past about −1.5 on Astia, because the base is already soft there and goes limp beyond it. A white point of 77.8 is reachable at −1.5 with exposure doing the rest.
• Color Chrome FX Blue Off, and this is the most load-bearing setting in the table. FX Blue deepens blues toward navy. Violet is 26.9% of this palette and it lives in the shadows, luminous rather than deep. Darkening it would turn the single most distinctive measured property into the thing every other painter here already has.
• Color Chrome Effect Weak rather than Strong. It acts on colours that are already saturated, and at chroma 11.9 there is very little for it to grip. Weak gives the few dense passages some gradation without thickening a palette whose character is lightness.
• Color 0, corrected down from +1 after the frames came out of the corpus. Measured chroma is 11.9, which is the second-lowest of twelve painters and well below Van Gogh's 18.6. Astia maintains saturation without help, and a positive Color here would take a deliberately gentle palette past what the reference actually does.
• Sharpness −1 and Clarity 0. A painting seen from across a room is soft, and the recipe softens slightly to match. Clarity stays at 0 for a honest reason rather than a stylistic one: broken brushwork is a surface property of paint, and a local-contrast control does not reproduce impasto. Pretending otherwise would be the easy mistake here. The side benefit is that this look survives HEIF intact.
• Exposure +1/3 to +2/3. A high-key frame with no black in it reads as overexposed to a reflective meter, which will pull it back down and hand you a grey version of the look.

Variants

Movie mode

Ports better than most here. Movie mode drops Grain, Color Chrome Effect and Clarity, and Clarity is already 0 while Color Chrome Effect is only Weak, so little of the design is lost. Astia, the white balance, DR400 and the tone curve all carry over, which means the high-key lift and the violet survive. 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 Astia and for each base rejected above, in this bank's Dynamic Range & Tone Curve, the contrast controls tone-curve table, and in a CIELAB measurement of fourteen public-domain reproductions against a five-painter control group.

No frames have been shot with it.

Test priority: in order

1. A sunlit subject with a real cast shadow, metered at +2/3. The whole recipe in one frame. Measure the shadow: the target is a black point near L* 35 with visible colour still in it. If the shadow is a grey hole, the recipe has failed at the only thing that distinguishes it.
2. Compare the shadow's b* against the highlight's b*. The target difference is about −2, against −9 to −24 for every other painter measured. This is the measurement the whole design rests on and the one most likely to come back wrong, because sensors and daylight do not behave like oil paint.
3. Anything violet or blue in shade. The Color Chrome FX Blue Off check, and at 26.9% of the palette this is the frame that matters most. The violet should read luminous rather than navy. If it has gone deep, confirm FX Blue is actually off, because X-Trans V renders blue deeper than X-Trans IV even at Off.
4. A pale high-key scene with no dark subject. The exposure test. At +1/3 it should read airy. If it reads grey, meter higher before touching the tone curve.
5. Foliage at midday. The yellow-green check, which is 30.3% of the reference palette and the easiest part to lose to a base that renders greens conservatively.

What would count as the recipe being wrong rather than needing a tweak: if the shadows will not hold colour at DR400 with Shadow −2, then a camera's shadow response is simply not a painter's, and the look has to come from shooting in light that is already open, which makes this a set of instructions about weather and time of day rather than a set of settings.

Pro tip: The whole recipe is one comparison. Every other painter measured has shadows 9 to 24 b* units less yellow than their highlights. Monet's gap is 2.2. If you take one idea to the field, it is that the shadow should be close to the same temperature as the light, which in practice means shooting when the sky is doing the filling.

Pro tip: There is no black in this look and there is not meant to be. A black point of 34.9 is the most lifted target in this bank by a wide margin. If your histogram touches the left wall, the frame is not this recipe, whatever the settings say.

Pro tip: Meter up, then up again. A frame with no black in it looks overexposed to the camera and it will correct you. Start at +1/3, go to +2/3 for anything pale, and +1 for snow.

Pro tip: FX Blue Off is doing more here than in most recipes. Violet is 26.9% of this palette against 0.0% for Rembrandt and Caravaggio, and it lives in the shadows. Deepening blue toward navy takes the single measured thing that separates this look from every dark-ground portrait recipe and throws it away.

Pro tip: Clarity 0 is an admission, not an oversight. Broken brushwork is paint sitting on canvas, and no local-contrast setting reproduces a surface. The recipe delivers the colour structure and leaves the touch to the painter, which is the honest division of labour.

HEIF-safe. Clarity sits at 0, so nothing is silently dropped. See Grain & Detail. Grain Effect, Sharpness, Clarity, High ISO NR.

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

Part 0: the idea, and the trap in it

Building a recipe from a painter is an appealing idea with an obvious failure mode: everyone already has an impression of what Monet looks like, and a recipe written from that impression would be a recipe written from a memory of a poster. This bank's method for aesthetics applies unchanged, which is to measure the corpus and let the numbers overrule the impression when they disagree.

The specific trap here is that a single painter's reproductions cannot be read on their own. Museum photography is colour-managed but not identical between institutions, and four centuries of varnish yellowing sits between an old master's palette and the file you download. Measure Rembrandt alone and you learn as much about varnish as about Rembrandt.

The answer is a control group. Twelve painters were measured with one instrument on one day, and only differences between them are treated as findings. Anything common to all twelve is left alone, because it could be the reproduction pipeline rather than the paint. That decision is what caught the errors in Part 1: a contaminated corpus moves one painter's numbers, and the control group is what makes the movement visible.

Part 1: the corpus, and a correction

Sixteen Monet oils from Wikimedia Commons, all public domain, all photographed cropped to the canvas, every one verified by eye. The control group is eleven other painters gathered the same way: Van Gogh, Rembrandt, Caravaggio, El Greco, Sargent, Titian, Vermeer, Klimt, Seurat, Botticelli and an attempted Turner.

This recipe was published once on a contaminated corpus and corrected the same day. The first harvest took reproductions on filename alone, which admitted four Monets photographed in their gilt frames. A gilt frame is large, bright and intensely yellow, so it inflated chroma, inflated the yellow share and diluted everything else. Two further problems turned up in the same audit and are recorded in the notes: Google Art Project serves drawings and sketchbook pages alongside paintings, and Caravaggio's full name is Michelangelo Merisi da Caravaggio, which quietly filled a Michelangelo folder with Caravaggios.

What the correction did to this recipe:

The base choice, the tone curve and the white balance all survived. One setting changed and one claim was walked back, which is recorded here rather than quietly amended.

The images were fetched into a scratch workspace and are not in this collection. They are public domain, so nothing prevented their inclusion. They are absent because this bank's standing rule is that reference images inform the work and do not ship with it.

Part 2: the measurement

Shadow and highlight are the darkest and brightest quintiles of each image, so the bands are per-image and relative rather than fixed, which keeps a dark painting and a bright one comparable. Vermeer (n=7 after deduplication) and Botticelli (n=5) are omitted from the table as too thin to carry a row, and Turner is omitted for the reason in Part 5.

Finding 1: there is no black, and the gap is larger than it first looked

Monet's black point measures 34.9 where Rembrandt sits at 4.7 and Caravaggio at 5.3. The old masters reach genuine darkness. This corpus does not approach it.

That is by a wide margin the most lifted target this bank has built toward. Reading Room, already described there as having a lifted black, measures 14.2. Specimen Plate is 13.7. Monet is more than twice either.

Finding 2: the shadows are close to the temperature of the light

The column asks whether the dark parts of a picture carry the same yellowness as the bright parts. The other eleven painters run from −9.34 to −24.34: their shadows are places where colour temperature drains away. Monet measures −2.24, four times closer to zero than the next painter and eleven times closer than Caravaggio.

A second estimator, using fixed absolute bands and a median in place of per-image quintiles, returns −3.04 for Monet against −20.42 for Caravaggio. The ordering survives a change of instrument, which is what this bank's Measuring film curves from datasheets, method, calibration, and its limits measurement discipline asks for.

The honest part. The first pass reported −0.17 and called it zero. That number came from a corpus with gilt frames in it, and the clean figure is −2.24 with a per-image standard deviation of 9.7. So this is a strong difference of degree rather than the absolute it was first written up as, and nine of sixteen Monets individually fall within three units of zero against zero of eleven Caravaggios. The claim that survives is that Monet drains far less colour from his darks than anyone else measured, which is still the reason this recipe exists.

Finding 3: the violet, which doubled

Blue-violet is 26.9% of coloured pixels, twice the next painter (Seurat at 13.3%) and absent in four of the eleven. It is the second-largest hue family in the corpus, behind yellow-green at 29.6% and ahead of yellow at 26.7%.

Taken with Finding 2, these shadows are violet and warm at once, which is a combination the camera has no single control for and which has to be assembled from a base choice and two refusals.

Part 3: the base

Four requirements, and Astia is the only simulation that meets them all.

1. A soft shadow end, because 34.9 is a very long way up and a base with a hard toe would have to be dragged there.
2. A soft highlight end, because 77.8 is a long way short of paper.
3. Saturation held at a gentle level, because 11.9 is the second-lowest of twelve painters and would flatten on a base that did not maintain it.
4. No per-hue agenda, because the palette runs yellow, yellow-green and violet in a specific relationship that the measurement found and a memory-colour base would rearrange.

Fujifilm's published description of Astia covers one, two and three in a single phrase: soft at both ends "while saturation is maintained". Requirement four follows from Astia having the least opinionated hue rendering in the set.

Why Eterna lost, since it is the obvious rival

Eterna is also soft at both ends, has twelve stops of latitude, and this bank's tone table specifically notes that Highlight −2 does not produce mush on it the way it would on a narrower profile. On tone alone it is arguably the better base.

It loses on one number. Eterna's blues are cyan-shifted by design, and blue-violet is 26.9% of this palette and the single most distinctive thing the measurement found. Pushing violet toward cyan would trade Finding 3 for a cast, and a base that is slightly better at the tone requirement is not worth a base that is actively wrong at the colour one.

Part 4: the zero that had to be argued for

White Balance sits at 0 Red, which is close to unique in this folder.

Almost every creative recipe here moves the Red axis, because almost every look has a cast to reproduce. The measurement says this one does not: shadow a* −1.14 against highlight a* −0.99, the same very slight cyan lean at both ends and no red at either. Warmth is mild and uniform, shadow b* +3.21 against highlight b* +5.45.

A Red shift would invent a cast the corpus does not have, and would work against the violet that Finding 3 identifies as the signature. The one notch away from Blue supplies the mild uniform warmth the numbers do show, and nothing else.

This is worth recording as a method point. The instinct when writing a warm-ish look is to reach for +Red, and here the evidence forbids it. A measurement is as useful for the settings it stops you making as for the ones it suggests.

Part 5: the honest limits

• No frames have been shot with it.
• The corpus is reproductions, so the endpoints are the endpoints of a photograph of a painting. Fit for targeting the appearance of the picture, which is what a JPEG recipe does, and unfit for any claim about pigment.
• Varnish and age affect the old masters in the control group more than they affect Monet, whose works are younger and less heavily varnished. That biases the control group warm and dark, which makes Findings 1 and 2 conservative rather than inflated: the gap would narrow, not widen, if every canvas were cleaned. Stated because it cuts in the recipe's favour and should be visible for that reason.
• Turner is absent from the control group and it is the one name most readers would expect. His painted output on the available open-access sources is swamped by the Tate's digitisation of his sketchbooks, and the medium filter could not separate them because the Tate catalogues those pages under painting categories. A measurement was taken and discarded: see the research file.
• Vermeer (n=7 after deduplication) and Botticelli (n=5) are too thinly sampled to carry a row and are excluded.
• The per-image spread on Finding 2 is large, standard deviation 9.7 for Monet. The between-painter difference is much larger than the within-painter noise, which is why the ordering is reportable and a single painting's figure is not.
• Broken brushwork is a surface property of paint. Clarity 0 is the recipe declining to pretend otherwise, and it means this look delivers the colour structure of the reference and none of its touch.
• A reference tells you what a look is and never whether a recipe reproduces it. Part 2 characterises the corpus and says nothing about whether these settings hit it.

Part 6: where it sits in the bank

Sixth Astia original, after Pastel Dream, Golden Hour Glow, Clean Girl, Cool White and Reading Room. It uses the base for its soft shadow end, which none of the others lean on this hard.

The pair to compare it against is Reading Room, because the two are opposite readings of what a lifted black is for. Reading Room lifts to 14.2 and puts red in the shadows with yellow in the highlights, a deliberate split. This one lifts to 34.9 and very nearly removes the split. Same instrument, same year, opposite conclusions, and both are in the bank because the corpora genuinely differ.

The nearest neighbour by subject is Soft Weather, which also works in open daylight without hard shadows. That recipe is for light that has failed, a milky overcast with the colour drained out of it. This one is for light that is working, where the sky is filling the shadows with blue and the sun is warming them back.

Last updated: 2026-09-17 (corrected same day against a clean corpus)