Free embroidery digitizer
Free embroidery digitizer that measures the artwork first
Drop artwork in and the first thing that comes back is a verdict with three counted figures behind it: how many colors it takes to cover nine tenths of the picture, how much of the working grid sits on a color boundary, and how much of it is thinner than 1.2 mm. Artwork that passes is planned as an angled tatami fill with satin on the narrow runs, a lattice underlay beneath and pull compensation on top, and leaves as a Tajima .dst. A photograph is given no file at all — the fill is drawn on request so you can see why, and that is the whole answer rather than a paywall. Free, no account, nothing sent anywhere.
- 100% free
- No signup
- Verdict before a file
- Satin decided per run
- Writes a Tajima .dst
Press Ctrl+V anywhere on the page to drop in a logo or a drawing from the clipboard — no need to click the field first.
How to find out whether your artwork can be embroidered
The measuring happens whether you ask for it or not; the work is deciding what to do about what it found.
Let it measure before you touch a control
The first thing on screen after the artwork loads is a verdict with the three numbers behind it. The colors needed to cover ninety percent of the picture separate flat artwork, which needs a handful, from continuous tone, which needs thousands. The share of the grid sitting on a color boundary says whether the design is mostly area or mostly edge, and thread can only fill area. The share thinner than 1.2 mm is the part no fill can hold at the size you have chosen. None of the three is an opinion about your picture; they are all counted off the working grid and they all move when you change the size.
Change the size before you change anything else
Almost every problem the assessment reports is a size problem wearing a different hat, because every feature in the artwork scales with the design. A shape 0.8 mm wide at 60 mm across is 1.6 mm wide at 120 and stops being a problem entirely — no density setting, no angle and no thread count can do the same thing. Move the width slider first and watch the narrow-area figure fall. Only when the verdict is green is it worth setting the stitch direction, and even then one angle covers the whole design, so pick the one that suits the largest area rather than the prettiest detail.
Read the four decisions it is not making
The panel at the bottom lists them by name: one stitch direction instead of a direction per shape, a sewing order taken from how much of the picture each color covers rather than from how the fabric will move, one type of underlay instead of one chosen for your cloth, and satin selected by run width along a single fixed axis rather than by each shape's own spine. Those four are the difference between this and a person with digitizing software, and none of them is a bug to be fixed later — they are what automatic means. Knowing which one is hurting your design tells you whether to adjust it here or hand it to somebody.
Technical specifications
| Artwork it measures | any raster the browser decodes — PNG, JPEG, WebP, GIF. Vector drawings have to be exported to pixels at the size you intend to sew |
|---|---|
| The three measurements | colors covering 90% of the picture, share of the grid on a color boundary, and share of the grid in areas under 1.2 mm across |
| When no file is offered | over 400 colors to cover 90%. Over 45% boundary or 35% narrow gives a warning and still writes the file |
| Stitch types it lays | angled tatami on wide runs and satin on runs under the threshold you set, classified per run rather than per shape |
| Underlay | a lattice at right angles to the top layer, at a quarter of its density, counted separately in the stitch table |
| Direction and compensation | twelve angles at 15° steps, one for the whole design; pull compensation 0 to 0.6 mm applied to the top layer only, never to the underlay |
| Size and stitch budget | 25 to 200 mm across, 1.5 to 4 rows per millimeter, 2 to 10 threads, and a hard stop at 60,000 stitches |
| What you can take away | a Tajima .dst when the verdict allows one, and a CSV stitch plan with the underlay, satin, fill and jump count for every thread |
Frequently asked questions
Why will this not digitize my photograph when other sites will?
Because the file those sites return does not work, and this one says so before taking your time rather than after. Thread has exactly one value per area — there is no half-tone, no opacity and no blending in a machine fill — so a continuous-tone image has to be cut into flat bands before it can be sewn at all, and the bands have no boundaries an algorithm can give a sensible direction to. What comes out is tens of thousands of stitches in arbitrary patches that read as mud from a meter away and take an hour on the machine. If you want to see it rather than take it on trust, the page will draw the fill for a refused picture on request, with no download attached, because it is a more convincing argument than this paragraph.
What is the difference between satin and fill, and why does width decide it?
Satin is one span of thread laid straight across a shape, side by side with the next span, which is what gives lettering and borders their glossy unbroken surface. Fill is the same idea broken into short stitches because the shape is too wide to cross in one. The switch between them is a physical limit rather than a style: a stitch much longer than about 10 mm has nothing holding its middle down, so it snags on a button, a finger or a washing machine and pulls into a loop. This page classifies every run it finds by its own width and satins anything under the threshold you set, defaulting to 6 mm, which is conservative and safe on garments.
What does underlay actually do?
Three jobs, and on a pile fabric the whole design depends on them: it holds the nap down so the visible stitches sit above it rather than disappearing into it, it pins the cloth to the backing so the top layer cannot drag it as the thread tightens, and it leaves the needle something firm to enter so an edge stays on the line it was drawn on. What this page lays down is a lattice — a coarse pass at right angles to the fill, at a quarter of its density. That is a sensible default and nothing more: a cap, a towel and a knit each want a different underlay, and picking between them is a judgement about the fabric in your hand that no page can make for you.
What is pull compensation and how much do I need?
Thread under tension pulls the fabric inward across the direction the stitches run, so a shape sews narrower than it was drawn and gaps open up between neighboring colors. Compensation cancels it by extending every run slightly at both ends before the file is written — 0.2 mm a side is a normal starting point on a firm woven fabric with cutaway backing, and a stretchy knit can want twice that. Too much is its own fault: over-compensated shapes overlap and the design looks swollen, with colors creeping over each other's edges. The honest way to set it is to sew a test on the actual fabric, measure the gap, and put half of it back on each side.
Why does making the design bigger fix so many problems?
Because thread is a fixed physical size and your artwork is not. A stitch is roughly 0.4 mm of thread lying next to another 0.4 mm of thread, and that never changes no matter how small you make the design — so at 50 mm across, a detail that takes up two percent of the width is one millimeter, which is on the edge of what a fill can hold, and at 100 mm the same detail is two millimeters and is comfortable. This is why commercial embroidery has conventional sizes and why a logo that looks fine on a jacket back falls apart on a cap front. The size slider is the most powerful control on this page for exactly that reason, and the assessment recalculates against it every time you move it.
Is automatic digitizing ever good enough to sell work from?
For simple flat artwork at a sensible size, yes, and it has been for years — a two- or three-color logo with clean edges at 70 to 100 mm, on a stable woven fabric with cutaway backing, comes out of an automatic fill looking like what a person would have punched. What it will not do is a script typeface at 15 mm, a design with a dozen colors touching each other, anything on a knit or a pile fabric, or a shape whose stitch direction carries meaning. The commercial rule of thumb is worth borrowing: if you would have to explain to the customer why it looks like that, pay somebody to punch it.
The preview looks denser than my artwork. Is that right?
Yes, and it is the underlay showing through. The layer beneath is drawn as a hairline in gray so it does not read at the same weight as the thread on top, but it is genuinely there and it is genuinely stitches — it is counted in the total, in the run time and in the thread. Turn the show-underlay box off to see only what will be visible on the finished piece, and turn the underlay itself off if you are sewing onto something firm enough not to need it. The stitch-type table keeps them separate on purpose: satin and fill are what somebody looks at, and the underlay column is what the machine spends time on before either of them starts.
About what digitizing decides, and which half of it can be automated
Digitizing is not tracing. Turning a picture into embroidery means deciding, for every area of it, four things that the picture itself does not contain. Which direction the stitches run, because thread is glossy and directional and a fill laid across a shape reads completely differently from one laid along it. What goes underneath, because a fill sinks into a knit and wanders on a pile unless something anchors it first. How much wider than drawn each shape has to be, because tension pulls fabric inward across the stitch direction and neighboring colors separate. And in what order the areas are sewn, because each one tightens the cloth slightly and the last color has to still land where the first one expected it. An automatic digitizer does the mechanical half — finding the areas, filling them, breaking runs at a sensible length — and guesses the four. This page makes the same guesses and then says out loud that it has made them, which is the only part of it that is unusual.
The reason a photograph cannot come out of that pipeline is not resolution and not effort. Thread has one value per area: there is no opacity, no half-tone screen and no blending in a machine fill, so a continuous-tone image must first be cut into flat bands of solid color. Those bands have no meaningful outlines, which means there is no shape to give a direction to and nothing for a satin edge to follow, and they interlock across the whole picture, which means the machine spends its time changing thread and jumping rather than sewing. What comes back is fifty thousand stitches that read as mud past arm’s length. Photographic embroidery does exist and it is beautiful, but it is made by a person choosing every stitch direction by hand, often with several thread weights and a lot of free-motion work, and it takes the time it takes. Any site that hands you a .dst from a family portrait in eleven seconds is selling you the version this page refuses to write.
Where that leaves a photograph is worth saying, because it does have an answer — just not a machine one. A counted chart can hold a photograph, because a person places every stitch individually and the medium has as many values as the palette does, which is what the cross stitch pattern maker is for. For flat artwork that only needs a fill and none of the assessment, the PNG to DST converter is the shorter road and puts the machine cost in front of you instead of the artwork measurements. And whichever of the two produced your file, the DST viewer will open it afterwards and count what is really in it — which is worth doing once, because a stitch count is the one number that predicts both what a design costs and what it will do to the fabric.
Product and company names shown here are the trademarks of their respective owners. Photo2Pattern is an independent tool and is not affiliated with, sponsored by or endorsed by any of them.
Where the artwork is assessed
Your picture is opened by JavaScript running in the tab you are reading this in. It is never uploaded, never sent to a server and never stored anywhere — there is no upload step to undo, because there is no upload. Close the tab and nothing of it remains.
The assessment is arithmetic over the pixels in this tab, so a design that belongs to a client can be measured here without it being handed to anybody — including us. Nothing about the verdict is stored, which is also why reloading the page starts the measurement over rather than remembering it.