Raster to vector for screen printing: turning a customer JPEG into plates

Why print shops end up vectorising

Almost nobody sends a print shop production art. They send a JPEG pulled off a website, a PNG exported from a phone app, a screenshot of a logo, or a design file at 72 dpi that somebody insists looks fine. Raster art is tied to the resolution it was made at: scaled up to print size the edges go soft, lettering picks up stair-steps on film, and JPEG compression noise that was invisible on screen becomes a fuzzy fringe on every edge once it is 12 inches wide. Vector has no such ceiling — the same paths drive a 4-inch left-chest print and a 14-inch back print with identical edge quality, and they output crisp film at whatever DPI your device runs. Vector is also the only thing a vinyl cutter, a Cricut or a laser can use at all, because those machines follow paths and not pixels.

Separate first, then trace — not the other way round

This is the part general-purpose vectorisers get wrong for print work. Point an auto-trace at a full-colour bitmap and it returns a stack of overlapping filled shapes ordered by whatever the algorithm found convenient. That looks right on screen and is useless as separation: you cannot burn a screen from "the shapes that happen to be visible in this region", because a screen is one ink everywhere it appears in the design, including the eleven letters of a wordmark and the two small highlights nobody thought about. Here the separation happens first, in colour space, with perceptual clustering — so an ink is one plate wherever it occurs — and then each flat plate is traced on its own. The plate is the unit. Each one comes out as its own SVG, in register with the others, which is a file a screen or a cutter can actually be driven from.

Clean the bitmap before you trace it, or the trace inherits the mess

Tracing quality is decided before the tracer runs, and this is where most of the disappointment lives. A JPEG that has been emailed around a shop four times carries compression noise along every edge, and a tracer will faithfully reproduce that noise as wobble in the path — thousands of tiny nodes describing artefacts rather than artwork. Two things help, and both are free and in the browser. Raise the tolerance so near-identical colours collapse into one ink instead of fragmenting a single edge into a dozen speckled plates. Then merge inks in the list until the plate count matches what you are actually going to burn or cut. If the art arrived on a background you do not want, remove it first — the same editor does that in a click — because a traced background is thousands of paths you will spend longer deleting than you saved by tracing. Look at the free raster plates before you pay for anything: if they look clean, the trace will be clean, and if they look ragged, no tracer will rescue them.

When to redraw instead

Sometimes tracing is the wrong answer and the honest move is to say so. A tiny logo with fine lettering — a 200-pixel-wide brand mark that needs to print two inches — has no edge information left to trace; a tracer will produce lumpy approximations of letterforms and the customer will notice. That job wants somebody to set the type in the actual typeface and rebuild the marks as real curves. Likewise anything you will have to modify structurally afterwards: a trace gives you paths, not an editable design. Tracing is at its best on flat art that is large enough and clean enough to have real edges — which is most band merch, most sports lettering, and most of what actually comes through a shop.

Photographs and halftones: do not trace them

Tracing a photograph is a bad plan. Continuous tone has no edges to trace, so the tracer invents thousands of them and produces a file that is slower, larger and less faithful than the raster it came from — the posterised look people sometimes want from this is better achieved by separating deliberately with a low ink count. The same applies to halftoned plates: a dot grid traced to vector becomes tens of thousands of tiny circles with no gain in quality and a file your cutter software will choke on. For tonal work use index or simulated process separation and keep those plates raster. That is not a limitation of the tool, it is the right answer.

What you get, and what opens it

One SVG per ink plus a manifest recording the ink colors and geometry, alongside the raster plates from the same render. The SVGs open in Illustrator, CorelDRAW, Inkscape and Affinity Designer, and drop straight into cutter software. If you would rather stay raster, the same render still gives you PNG plates and film output — the vector trace is an addition to the deliverable, not a different product you have to choose between. And separately from the render, free in your browser, you can export a multichannel PSD carrying one real spot channel per ink.

Free separation, paid trace

Load the art and separate it in your browser for free, with no account and no upload — you will know within seconds whether the plates are clean, which is the part worth checking before you spend anything. The vector trace runs as part of the server-side print-ready render at 10 credits per image, with free re-renders of that image for 24 hours while you adjust the ink count or tolerance and re-trace. Credits come with a monthly plan, or as a one-off pack whose credits never expire — see the pricing page for current prices in your currency. No watermarks on any output, ever.

It also unlocks the cutting work

Layered vector output is not only a screen-print deliverable. The same per-ink SVG plates are exactly what layered vinyl, Cricut and laser projects need: one path set per material color, in register, ready to cut. If you came here from cutting rather than printing, the separation logic is identical — you are just calling the inks materials, and the choke control that stops an underbase peeking out is the same control that stops two layers of vinyl fighting for the same millimetre.