Index color separation, and when to choose it over simulated process

Why index is the quiet favourite

Conventional tonal printing builds a gradient from dots of varying size, which forces you to pick screen angles that do not interfere, compensate for how much each dot will grow on press, and hope the fabric agrees. Index separation sidesteps most of it. Every dot is the same size and 100% ink; tone comes from how densely those dots are scattered. Because there are no varying dot sizes there is no angle set to choose and no rotated dot grid for a neighbouring plate to beat against, so moiré stops being the failure mode you spend your afternoon managing. And because every dot is solid, dot gain moves the image slightly darker instead of destroying a tonal curve. It is forgiving on textured garments, forgiving on old equipment, and forgiving of a printer who is still learning tonal work.

How it works here

You choose the ink count. The image is clustered in LAB space to find the palette that best represents it — the same perceptual clustering the spot separator uses, so the inks it picks are inks a person would have picked. Then the transitions are dithered with blue noise rather than an ordered pattern, because ordered dithers leave a visible grid that fights the weave of a shirt. Bump the ink count and the image gets smoother and the screen count goes up; drop it and you get a graphic, posterized look that a lot of merch actively wants.

One thing to know first: the free engine keeps the biggest ink as the base

This is the sharp edge on the free path and it is easier to hear now than to discover at the light table. The free in-browser engine always treats the largest-area cluster as the base and leaves it in place rather than lifting it to a plate, on the theory that the biggest flat field in a design is the garment or the field the art sits on, and you do not burn a screen for the shirt. On flat artwork that is exactly right, which is what it was written for. On a photograph it is a guess, and it is the kind of guess a photograph punishes: if the largest area of your image is a sky, a wall or a subject's jacket, that colour becomes the base instead of a plate — and it is a colour you almost certainly do need to print. So the free index separation of a photograph hands you one fewer plate than the ink count implies, and the missing one may not be the one you would have chosen. Look at what landed in the base before you trust the plate list; if it is not your garment colour, treat that plate as one you still have to print. The paid render decides this differently and more carefully: it drops an ink to base only when that ink is genuinely close to the garment colour you set — within a small perceptual distance — and it tells you in the render warnings which ink it did that to and why. That is why free and paid index separations of the same photograph can legitimately come back with different plate counts. It is this difference, not a bug. On flat art where the biggest area really is the shirt, the two agree.

What you can control

Ink count from 2 to 12. Tolerance, which decides how eagerly near-identical colors collapse into one ink. And the full ink list — merge two inks that are practically the same, rename a plate to the ink you will actually mix, reorder the print sequence, or delete a plate and let the rest absorb it. Per-plate preview shows each screen on its own before you burn anything. Garment colour is a control too, and it is worth being exact about what it does at each tier: on the free client-side path it tints the on-screen proof so you are judging the composite against the shirt you will actually print rather than against white — it does not change the plates themselves. On the paid render it genuinely feeds the separation: the underbase is built for that substrate, and an ink close enough to the garment colour is recognised as the substrate rather than cut as a screen. So set it either way — on the free path for an honest proof, on the paid path because it changes the output.

Free, in the browser, with no upload

Index separation is client-side, like spot. No account is needed, and the separation makes no network call at all — your artwork is never uploaded, which you can confirm in the browser's network tab. Export per-ink PNGs, a numbered ZIP, an ORA file, a layered PSD, or a multichannel PSD with one real spot channel per ink (one PSD export a day on a free account, full quality, no watermark ever; uncapped on a monthly plan). Because index plates are already solid dots rather than continuous tone, the free export is genuinely usable film artwork for a lot of jobs — which is not something we can say about a tonal separation.

When you still want the paid render

Three reasons. First, resolution: dithering is resolution-dependent, and a plate rendered at 1200 dpi for a 14-inch print is a 16,800-pixel image — comfortable on a server, painful in a browser tab on a shop PC. Second, print math: a choked underbase under an index separation on a dark garment, per-ink choke and spread, registration marks, and vector or 1-bit film output. Third, the base decision above — the paid render only demotes an ink to the substrate when it actually matches your garment colour. That render is 10 credits per image, and for 24 hours afterwards you can re-render that image as often as you like, free, for 24 hours.

Index vs simulated process: what actually differs

Both print a full-colour image with spot inks, and they differ in one structural way from which everything else follows. Index makes every dot the same size at 100% ink and builds tone out of dot *density*; simulated process varies dot size and builds tone out of dot *area*. Because index dots never change size there is no dot grid to rotate, so index needs no screen angles at all and there is no rotated grid for the plate next to it to interfere with — moiré is not the failure mode you spend the job managing. Simulated process has to halftone every channel, which means an angle set, and moiré is a live risk you manage. Against that, varying dot sizes carry more information: simulated process resolves more tonal detail and reads smoother, particularly in skin tones and long soft gradients, where index shows its dither pattern and its palette steps. Index also degrades gracefully — dot gain moves an index print slightly darker overall, because a solid dot growing is still a solid dot, whereas gain in a halftone distorts the tonal curve and is the classic cause of a muddy process print.

So which one should you run

Choose index when the image will tolerate a graphic look, when the garment is textured or heathered, when your equipment or your registration is not razor-sharp, when you are still learning tonal work, or when you want the job to behave the same way on the fiftieth shirt as on the first. Choose simulated process when the artwork is photographic and needs to look photographic — faces, chrome, soft product shots, anything where a visible dither pattern would read as a mistake rather than a style — and when you have the press control and the patience to correct channels. There is also an honest cost difference: index separation here is free and runs in your browser, and simulated process needs a server render and a human on the channels, because fully automatic simulated process is unreliable in every tool that sells it, ours included. Plenty of good shops run index for most of their tonal work and reserve simulated process for the jobs that genuinely demand it. Try both on the same file — switching separation type keeps your document, so it costs you a click rather than a restart.