Reading Room

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

---

Part 0: what the source says, and a correction to our own notes

The reference is the Consumer Aesthetics Research Institute's entry for Frasurbane, named by Evan Collins, running from the late 1980s to the mid 1990s.

This bank's internal queue carried an error about the corpus that had to be fixed before any work could start. It recorded the entry's Are.na channel as holding 778 blocks, "by far the biggest of any entry in this queue", almost all untitled 2018 screen captures, and concluded that the written entry carried more information than the image index did.

Two channels by the same researcher had been conflated. The 778-block channel exists and is the one full of untitled screenshots. The channel CARI actually links from the entry holds 36 blocks, every one titled and dated, which is comparable to Austurbane's 24 rather than the largest in the file. The queue's conclusion followed from the wrong channel, and it was backwards: for this look the image corpus carries considerably more than the written entry, because it can be measured.

Part 1: the corpus, and the decision to measure it

Thirty of the 36 blocks are photographs. The other six are a type specimen, two interface screenshots, an illustrated album cover and a video-noise frame, and they are excluded from every number below. That exclusion list was itself corrected once during the work: a first pass dropped a real photograph and kept the video frame, which is worth recording because it changed the aggregate before it was caught.

The images were fetched into a scratch workspace for analysis and were not downloaded into the reference set, are not reproduced here, and are not in this collection. CARI's notice is that its collection is a source of inspiration and analysis rather than free assets, and this bank is built to be sold, which makes that binding. The measurement travels. The pictures do not.

Part 2: the measurement

Each photograph was converted to CIELAB and split by luminance. The question was simple and the answer was not the expected one: where in the tone scale does the warmth live?

The three estimators are structurally different. The first uses fixed absolute band edges. The second uses per-image relative edges and a median, so it is immune both to the choice of edge and to outlying pixels. The third uses no bands at all, regressing chroma directly on luminance across every pixel. This bank's Measuring film curves from datasheets — method, calibration, and its limits curve-measurement discipline holds that a finding is only reportable if it survives a change of estimator rather than a change of window, and this one does.

The sign tests are what make it convincing rather than merely consistent. Shadow a* is positive in 29 of 30 images. The b* slope is positive in 24 of 30. Shadow b*, by contrast, is positive in only 20 of 30, which is barely distinguishable from a coin flip, and that near-null is the finding: there is no yellow in these shadows to speak of.

The warmth is split. Red sits low, yellow sits high, and the two run in opposite directions across the tone scale.

The endpoints matter too.

One honest limitation. These are scans and screen captures of printed material, so the endpoints are the endpoints of a reproduction rather than of a negative. That would be fatal if the goal were to characterise an emulsion. It is not: the target here is the appearance of the finished page, which is exactly what a JPEG recipe is trying to produce. The instrument matches the question.

Part 3: why the warmth is split, physically

The structure is not an arbitrary grade. It is what a tungsten-lit room does to a camera that has not fully corrected for it.

The lit surfaces in these frames are lit by tungsten practicals, so the bright parts of the image are the parts carrying the illuminant's yellow. The dark parts are lit by bounce and fall-off, where the tungsten spectrum is both weaker and, after the era's print stage, biased toward the red end. The result is yellow where the light is and red where it is not.

That reading is testable, and it was tested. Modelling a white surface under a tungsten illuminant, captured with the camera adapted to a given Kelvin value and viewed under D65, reproduces the measured highlight chroma when the camera setting sits about 400 K above the illuminant.

Target: a* +2.69, b* +14.23. The three closest fits are all offsets of 400 K, across three different illuminant temperatures, which is a stronger result than a single match would be. A daylight setting overshoots by roughly a factor of four.

Two consequences for the recipe follow directly. Auto white balance is an active failure here rather than a safe default, because its whole job is to remove the cast the look is made of. And the correct instruction is an offset rather than a value, because the look is the difference between the light and the setting.

Part 4: the base

Four requirements come out of Part 2, and they are unusually specific for a creative look.

1. Soft at both ends. L* 14.2 under L* 87.9 is a curve with no hard shoulder and no hard toe.
2. Saturation maintained. C* 18.9 is moderate. Not a desaturated look, not a saturated one.
3. Tolerant of uncontrolled colour temperature, because that is the entire lighting condition.
4. No competing per-hue signature, because the recipe needs to place red low and yellow high without arguing with the base about it.

Astia satisfies all four, and it is the only one that does. Fujifilm publishes it as soft at both ends "while saturation is maintained", and engineered for mixed and uncontrolled colour temperature. Requirements one, two and three are the manufacturer's own description of the simulation. Requirement four follows from Astia having the least opinionated hue rendering among the candidates.

Why Nostalgic Neg. Lost, since it was the queue's hypothesis

The queue proposed Nostalgic Neg. With a confident argument: its amber is anchored in the shadows, warm sepia is amber, and the one property that made it wrong for Austurbane makes it right here.

The measurement is what defeats it, and the defeat is precise rather than approximate. Nostalgic Neg.'s amber is red and yellow together, low in the tone scale. The corpus shadows carry a* … and b* …. The red is there. The yellow is not. Nostalgic Neg. Would deposit yellow in exactly the band this look keeps clear, and Fujifilm states its amber "does not change easily when adjusted via white balance alone", so there is no dialling it back out afterwards.

This is worth recording as a method result, over and above the base decision it settled. The hypothesis was reasonable, it was arrived at from the written entry, and it was wrong in a way no amount of reading could have exposed. It took thirty photographs and a colour-space conversion. The queue's own instruction for this look, that the images have to be looked at rather than parsed, turned out to be right for a better reason than the one it gave.

The rest of the field

Part 5: how two global controls make a split-tone look

The camera has no split-toning control. It has a Kelvin value and a two-axis shift, and both are global. The look is nonetheless reachable, for a reason worth spelling out.

The Kelvin offset supplies yellow. Because the yellow arrives with the illuminant, it lands on the surfaces the illuminant is actually lighting, which are the bright ones. Nothing about this is a tone-curve trick. It is the subject doing the work, and it is why this recipe is more dependent on shooting the right room than most in this folder.

The WB Shift supplies red. Fujifilm's grid runs Red against Cyan and Blue against Yellow, with no green axis, so Red and Blue together produce a magenta nudge: red with the yellow taken back out. Applied globally, a fixed magenta bias is plainly visible where chroma is low and effectively invisible where chroma is already high. Shadow chroma in this corpus is low. Highlight chroma is b* …. The same absolute bias therefore reads in one band and vanishes in the other.

… and … is small on purpose. The shadows need roughly five units of a* beyond what the Kelvin offset already provides, and the highlights must stay near a* …. A heavier shift reddens the highlights and collapses the separation.

This is the part of the recipe most likely to be wrong in the field, and the notes says what would falsify it: if the shadows will not go red without taking the highlights with them, the mechanism is too weak and the split has to come from the lighting instead.

Part 6: the settings that follow

• … is chosen by a measurement rather than by habit. The highlight that has to survive is a yellow one at b* +14, and a clipped highlight is neutral white. Two stops of protection keep the most distinctive measured feature alive. The usual objection, that DR modes lift everything below the highlights, is not a cost here, because the target black is lifted anyway.
• Highlight … and Shadow … are … pushes on a base that is already soft at both ends. Shadow stops at …ecause … is lifting too, and the pair should land near L* 14 rather than overshoot into milk.
• Clarity … stands in for soft focus, which the CARI entry names twice and which a recipe cannot actually deliver, because it is a lens and diffusion decision. Clarity is a local-contrast operator and moves in the right direction. It is not the same thing, and the recipe says so rather than pretending.
• Sharpness …, not lower. Wood grain, laminate, stone and the tooth of textured paper are literal subjects in this corpus. Softness belongs at the macro scale.
• Color Chrome Effect … acts on already-saturated colours by giving them tonal separation. In a frame of mahogany, brass and oxblood on a soft base, that is the exact problem it exists to solve.
• Grain … and …, because commercial print and transparency are clean stock. Enough for a print surface, not enough to claim fast film.

Part 7: where it sits in the bank

This is the fifth Astia recipe, after Pastel Dream, Golden Hour Glow, Clean Girl and Cool White, and it uses the base for a property none of the others touch: tolerance of an uncorrected cast rather than softness or high-key rendering.

It pairs with Austurbane by construction, which is why the queue ranked them together. Same staged-object subject, opposite decade, opposite temperature, and as it turns out opposite base: Classic Chrome's cool shadows against Astia's neutrality, hard tone against soft. The pair is more useful now than when it was proposed, because the two recipes disagree about almost everything.

Three neighbours it should not be confused with. Revival House is warmer and a decade earlier, and its warmth is uniform rather than split. Clay Season is warm by hue but restrained in saturation, where this one keeps chroma at 18.9. North Light is the neutral interiors recipe, which is what this becomes if you correct the white balance properly.

Part 8: honest limits

• No frames have been shot with it. Everything above is a target derived from other people's photographs plus a colour-adaptation model.
• The endpoints are measured from reproductions of print, not from negatives. Fit for a JPEG recipe, unfit for any claim about an emulsion.
• The 400 K offset is the one number that came from a model rather than from a photograph. It is the most likely thing in the recipe to be off, and test 2 in the notes is designed to check it directly.
• A reference tells you what a look is and never whether a recipe reproduces it. The measurement in Part 2 characterises the corpus. It says nothing about whether these settings hit it, and only test shots can.
• Soft focus is not delivered. Clarity … moves in the right direction and is not a diffusion filter.

Last updated: 2026-09-08