Under the hood
How PiperStitch actually digitizes.
For the curious: what a digitizer decides, how PiperStitch decides it for you, and how we know it works. No jargon without a translation.
First, the big idea
Rules, not guesses.
An embroidery machine doesn’t sew pictures. It sews a list of needle positions, in order, with a thread-colour change here and a trim there. Turning a picture into that list is called digitizing, and it’s a craft: a good digitizer looks at a shape and decides what kind of stitch it wants, how dense, in which direction, what goes underneath it, how much to over-size it so the fabric’s pull doesn’t shrink it, and what order everything should sew in so the needle never drags a visible thread across your design.
Cheap “auto-convert” tools skip all of that. They trace the pixels and fill everything with the same stitch, which is why the results pucker, gap, and look nothing like the picture. PiperStitch is different in a specific way: it makes the same decisions a professional digitizer makes, using the same rules they use, one shape at a time.
It’s worth saying plainly: PiperStitch is not an AI model. Nothing here was “trained” on a pile of designs to guess what stitches might look right, and nothing is sent to a server to be generated. It’s an engine of explicit, written-down digitizing rules — the same ones taught to human digitizers — applied to the measured geometry of your artwork, on PiperStitch’s own servers. That means it’s predictable: the same artwork gives the same result every time, every choice has a reason you can see in the Inspector, and you can override any of them.
The three stitches
Every embroidery design is made of just three things.
Choosing the right one for each shape is most of the craft. PiperStitch measures every shape and chooses for you — then re-chooses if you resize.
Running stitch
What it is: a single line of stitches, one after another, like hand sewing. A bean stitch sews each step forward-back-forward for a bolder line.
When PiperStitch uses it: outlines, fine details, and anything narrower than about 1.5 mm — too thin for satin to hold.
Satin
What it is: long stitches zig-zagging edge to edge across a band. The classic glossy look of embroidered lettering and borders.
When PiperStitch uses it: anything from 1.5 mm to about 12 mm wide. Wider than that and the long stitches snag, so it switches to fill — or splits the shape and satins each part.
Tatami fill
What it is: rows of short stitches, like a woven mat, covering a large area. Each row’s stitches are offset from the last so no grid pattern shows.
When PiperStitch uses it: big shapes and backgrounds, and any satin candidate that turns out too wide. Rows run across the shape’s narrow direction so they lie flat.
Step by step
What happens when you click Create Embroidery File.
Your picture becomes shapes
PiperStitch first works out how many colours your design really has. A two-colour logo saved as a JPEG contains thousands of slightly different pixels; PiperStitch finds the two that matter. Each colour region is then traced into a clean outline — with its holes kept as holes (the middle of an O stays open) and overlapping regions separated into layers, so a badge doesn’t lose the disc behind its text. If you give it an SVG, it uses the exact vector paths instead.
Each shape gets measured and classified
The width of every shape is sampled at a dozen points along its length. Narrow → running stitch; medium → satin; wide → fill. A whole word of lettering is judged together so every letter sews the same way, and letters with counters (O, P, R, A, D, Q) get a satin ring around the hole rather than being flattened into fill. Resize the design and this is all reconsidered — scale small text up and it becomes satin on its own.
Underlay goes down first
Professionals never sew satin straight onto fabric: a light preliminary layer — underlay — stabilises the cloth and gives the top stitches something to sit on, which is what makes satin look raised and even. PiperStitch lays an edge-run underlay under every column and, for wider columns, adds a zig-zag layer too (the technique digitizers call “German underlay”). Fills get their own edge-run and a light cross-hatch.
The shape is over-sized to fight pull
Thread under tension pulls fabric inward, so a satin column always sews a little narrower than it was drawn, and a fill pushes outward at its ends. Digitizers counter this with pull compensation: drawing the shape slightly bigger than they want it to end up. PiperStitch does this automatically, scaling the amount to stitch density and to the fabric you picked — twill barely moves; a stretchy knit or a plush towel needs more.
Stitches are generated, with care at the edges
Satin density tightens automatically on the inside of curves and relaxes on the outside so coverage stays even round a bend. Column ends are squared off the way a hand digitizer squares them. Fill rows are staggered so seams never line up. Every stitch is then checked: too short to sew (under 0.15 mm) or too long to be safe (over 12.5 mm) and it’s split or merged. Each object starts with a tie-in and ends with a tie-off so nothing unravels.
Everything is put in order
Backgrounds before foregrounds; inside before outside; the first letter of a word first, not whichever pixel a scanner happened to reach. Then the order is refined to keep the needle’s travel short, and any travel that can’t be avoided is routed underneath stitching that comes later — hidden travel — instead of scratching across your design. Colour changes are grouped so your machine stops as few times as possible; long jumps get a trim.
It checks its own work
The Embroidery Readiness report scores the result from 0 to 100 and lists every real problem it found, with the actual numbers: a satin column too thin to hold, detail too fine for the fabric, a jump longer than 15 mm, a design that won’t fit the hoop, an unusual number of trims. Never “density problem” — always what, where, and what to do about it.
How we know it works
Built from the craft. Tested against real thread.
Professional digitizing practice, written down
Every threshold and technique in PiperStitch comes from documented digitizing guidance — the width limits for satin, underlay types, compensation, sequencing — and from studying the two most respected open-source embroidery engines line by line. Where we went beyond them (automatic fill angles, curvature-aware satin density, reading-order sequencing) it’s our own work, and it’s documented.
290 automated checks, on every build
Every export is written and then read back by a separate parser — and, in development, by an independent embroidery library maintained by another team — and compared stitch for stitch. Real files from other manufacturers’ software are opened and checked too. If any of it disagrees, the build fails.
Real sew-outs changed the defaults
Designs are stitched on a real Brother machine and compared against the preview. That’s how the default density became 0.32 mm instead of 0.40 (the wider setting showed rows), how sew order came to start at the first letter, and how the pull-compensation table got its fabric-specific numbers. Real customer logos found real bugs, and each one became a permanent test.
You’re still the digitizer
Automatic isn’t the same as locked.
Most of the time you’ll click once and export. But every decision above is exposed, per object, in the Inspector: change a shape from fill to satin, set its density, its underlay, its fill angle, its compensation. Merge shapes, paint in coverage, erase a fragment, reorder the sew sequence, move and resize on the canvas, choose threads from a real manufacturer palette, and watch the realistic preview and the readiness score update as you go. The engine does the first 95% so you can spend your attention on the 5% that makes a design yours.
See the decisions on your own artwork.
Free for 14 days, every feature. Open the Inspector and watch what it chose — then change your mind.