#Overview

A single negative has to fit an image's whole tonal range into what one ink can render. Sanguine relaxes that constraint: Calibration Flow splits the negative into 2 or 3 partial negatives, each meant for a different ink — yellow carries the detail in the highlights, burnt sienna the mids, red the blacks. The plates are exposed one after another, in register, and the stack rebuilds the complete image.

The principle is a relay race: each ink runs its leg, and the legs add up to exactly the full distance. Before a single file is produced, the app verifies that reconstruction numerically; if the sum drifts from the source image, nothing is generated and a message says so. You cannot expose two or three sheets only to discover at print time that they do not recompose.

What the app does not decide. One ink layer per plate, or several plates on a single layer: both arrangements exist, and the maths is the same either way. Your process settles it, not the software.

#Choosing the number of inks

The option lives in the Type tab, under Negative and Positive: 1 negative, 2 inks or 3 inks. By default the app stays on 1 negative — the classic single image. The split works in both polarities; it makes no sense in CMYK, Colour or Full Gamut, where each channel is already a separate plate. With 3 inks the range is shared between yellow, burnt sienna and red; with 2, burnt sienna is the one that goes — never red, which closes the scale. The preview regenerates on every change and shows the plates side by side, in a horizontally scrolling row.

Each plate is identified by its label — its exposure rank and the bounds of its slice: "Negative 2/3 — 42-77 %". The label names no ink, and that is deliberate: which pigment goes on which plate is a darkroom decision, yours, and it can change from one print to the next. The rank never changes — it is the order in which the transparencies are exposed.

That label becomes the filename of the download, along with the set's timestamp:

negatif2sur3_42-77pct_20260824-2130.png

. Files from one set stay grouped and sorted in exposure order, and the same set downloaded in English or French gives exactly the same names. The Save button downloads the whole set at once; each card also keeps its own download button.

#Setting the limits yourself (advanced)

By default, where each ink stops comes from the method: an uneven split, tighter towards the shadows. The reason is physical — alternative processes compress in the blacks, where every extra second of exposure buys less and less density. A narrow slice buys back the separation the process loses there, because the ink spreads its whole range over that slice alone.

That split is a starting point, tuned for one family of processes. Yours has no reason to compress in exactly the same way, and your calibration curve — applied before the split — has already corrected part of it. Under Advanced settings you can therefore set where each ink stops, in percentage of ink.

Three things to know:

You only enter the intermediate limits. The last ink always closes the scale: with 2 inks there is one limit to give, with 3 there are two.
An empty field is not a zero: it means "the method's values". The Method's values button clears the fields, it does not write numbers into them.
An even split is not neutral in a thick process. It gives every ink the same fineness, but in a process where the layer has thickness — gum, carbon — even slices produce relief of uneven heights, and the tallest points can collapse in the shadows. Tightening the dark slices is precisely what avoids that. Make the trade knowingly, not by default.

Rejected limits are refused immediately, with the reason: they must increase, stay above zero and below one hundred. Nothing is generated until they hold — never a plausible, wrong set of plates.

This is the setting that decides your registration tolerance. It lives under Advanced settings, below the limits, and is given as a percentage — 0 to 100. Its two ends need no explaining:

At 0 %, each ink carries only its own slice. This is strict separation, the setting that gives the split its full benefit. It is also the most fragile: a stacking error puts one ink where another belonged, across a whole zone of the image, not just along an edge.
At 100 %, the three negatives are identical. Separation disappears entirely — each ink spreads its full range over the whole image again — but there is nothing left to misalign: all three inks are everywhere in the same proportion. A registration error can no longer swap one colour for another; it leaves only a slight edge doubling, as in any colour print. Exposure times then divide in proportion to the slices.

In between, the slider is a direct, legible trade: how much separation you accept to lose in order to buy registration tolerance.

Three things to know:

Zero is a setting, not an empty field: it is strict separation. An empty field means "the engine's value".
The whole range is useful. The setting acts on the entire image at every position, and reconstruction stays exactly verified from one end of the travel to the other: nowhere does the slider break the sum.
At 100 %, your limits no longer matter. Since the negatives are identical, the split plays no part. There is no need to tune both settings together at the top of the range.

To choose a position, the safest move is a dry run: print the registration negatives alone, stack them on a scrap of paper under your real working conditions, and look at the displacement you actually get. Genuinely good registration allows a low sharing value; approximate registration is bought back by raising it.

A word on how this relates to the split: tightening the dark slices protects the relief from collapsing, but gives you no registration margin at all. They are two distinct settings and do not substitute for each other — sharing is the only one that buys tolerance.

#Exposure time is the scale factor

The app computes the shape of each plate; the amount of light each one receives is your darkroom decision. Each ink's exposure time is found with a test strip, as always in alternative processes — it is what sets each ink's real density. A dark ink exposed longer deepens the blacks; a light ink exposed shorter protects the highlights. The app neither imposes nor stores these times: they describe your process, not your image.

#The targets: the large one rebuilds, the small ones line up

One point of physics governs everything else: the plates print one on top of another, at the same place on the sheet. Two marks placed at the same point on two plates therefore deposit material there twice. Repeating an identical wedge on all three plates would not give a wedge three times more reliable: it would give a blot.

The targets account for this, each in its own way. With registration marks enabled, every plate carries:

The large bottom wedge, which follows the split. Each plate shows its own slice of the range, at the same place as the others. The deposits add up, and the complete wedge appears on the paper: that is what you read at print time to judge calibration, exactly as on a single negative. If it comes out hollow or saturated, it is showing you your exposure times.
The small top wedges, one per ink, each in its own slot. With 2 inks the top margin splits in two, with 3 in three — and each plate occupies only its own, numbered. They never overlap: after stacking, the two or three small wedges appear side by side, and your print's composition reads at a glance, ink by ink.

The control strip (5×5 grey scale) and the zone targets follow the same logic as the large wedge: they pass through each ink's table like the image itself, and rebuild on stacking.

#Works with zones, curves and registration marks

The split combines with curve calibration (the curve is composed into each plate before the split), with multi-zone correction (each ink's table is composed zone by zone, with that zone's own curve) and with registration marks, which are essential here: the four corner targets and the orientation key are what guarantee that successive exposures land on top of each other exactly.

#In practice

Type tab → Negative (or Positive) → choose 2 or 3 inks.
Check the preview: the plates appear in a row, each carrying its rank and its slice.
Turn on registration marks — they are what makes the stacking exact.
Download the set and print each plate on its transparency.
Establish each ink's exposure time with a test strip.
Expose the inks one after another, in register, in file order.

The stacking order is settled by trial, on two criteria that can contradict each other: how transparent each ink is, and how light or dark it is — the darkest usually going last. When in doubt, print a test overlap both ways round and judge by eye.

The feature is free, on the web and on iOS.