The word makes a promise the process does not keep. Digitizing sounds like converting — something analog becomes something digital, a picture becomes a file, and the interesting part is which format comes out the other side.
What actually happens is that somebody writes a sequence of instructions for a machine, using the picture as a description of the intended result.
Digitizing is authoring, not converting
A conversion preserves content in a new container. A photograph saved in another format is the same photograph; nothing had to be decided.
Digitizing decides a great deal. Stitch types, the direction each area runs, densities, the underlay laid down before the visible stitching, the order everything happens in, how the needle travels between areas, and compensation for the way fabric distorts while it is being stitched. None of those are present in an image, none can be derived from one, and together they account for most of what the finished piece looks like and how long it takes to make.
The misconception costs money because it decides who is treated as responsible. If digitizing is a conversion, the person doing it is a vendor performing a mechanical service, the brief is the artwork, and the output is judged by whether it matches. If it is authoring, they are making design decisions on your behalf, and the brief needs to contain the things you have opinions about. Teams holding the first model brief almost nothing and then argue about the result, which is an argument about judgments that were never requested and could not have been guessed.
This picks up exactly where the artwork side leaves off. What the person supplying the file controls — shape structure, point counts, colors, whether regions share an edge — is a separate set of decisions with its own treatment: what a vector commits you to before it reaches a machine. Everything below belongs to the step after that one.
What is actually in a stitch file
A stitch file is not a picture of embroidery. It is a list of positions and commands, worked through in order: move here and stitch, jump to there without stitching, trim, stop for a color change, resume.
That structure explains several things that puzzle people:
• The file has no resolution, because it contains no pixels — only coordinates and instructions.
• It is specific to a size, because the coordinates are absolute rather than relative to a shape.
• It carries an order, which an image cannot, and the order affects both cost and appearance.
• It may or may not carry color information, depending on the container, because color in embroidery is a thread selection made at the machine rather than a property of the geometry.
Which brings up the format question, usually asked as a request: which file type do you need. That question has a short answer available from whoever is running the machine, and it is not the distinction that matters.
The distinction that matters is between an editable digitizing source and a machine file. The source holds the objects, their assigned stitch types, and the parameters behind them, and can be revised. The machine file is a rendering of that source into a flat sequence — like a flattened output next to a layered original. Keeping only the machine file means every future change is a fresh digitizing job rather than an edit, and this is the single most common way a company pays for the same work twice. Asking for and storing the editable source alongside the machine file is worth doing at the point the job is commissioned, because afterward it may no longer exist: where both files should come from.
Direction is a decision the image never contained
Thread is shiny, and stitches lie in a direction. Light striking a stitched area reflects according to which way the threads run, so the same shape filled in one direction and filled in another reads as two different tones of the same color.
This is not a subtlety visible only to specialists. It is the mechanism behind most of what makes embroidery look considered or look flat — a leaf whose fill follows its own curve reads as a leaf, and the same leaf filled in one uniform direction across the whole shape reads as a sticker of a leaf.
An image has no direction field. There is nowhere in the artwork to say which way an area should run, so whoever digitizes it decides, area by area, and those decisions are visible in the result. A designer who cares about that outcome is depending on somebody else’s judgment unless they say something, and saying something is cheap.
Sequence is a decision too
Order is the other piece of information an image cannot hold, and it does more work than most people expect.
The sequence determines how the needle travels between areas, how many times thread has to be trimmed, and how many times the machine stops. It also determines how much the fabric has been pulled around before any given shape is laid down — an area stitched late sits on cloth that has already been distorted by everything stitched before it, which is why two files containing identical shapes can produce pieces that register differently against each other.
The symptom people report is registration — two elements that should meet do not quite meet, or an outline sits slightly off the shape it is outlining. That gets blamed on the artwork, where the two elements are perfectly aligned, and the artwork is then redrawn to no effect. The misalignment was introduced by the order in which the cloth was worked, and the correction belongs in the sequence rather than in the drawing.
A design with many separated small elements is expensive for this reason rather than for its stitch count, since the machine spends its time traveling and trimming. That cost is created during digitizing and is largely invisible when looking at the artwork.
A good stitch file does not match the artwork
Fabric is pulled inward by stitching. A digitizer compensates by making shapes slightly larger than drawn in the direction the pull will act, so that after the cloth contracts the result lands where the design intended.
The consequence is worth stating plainly, because it inverts the obvious check. A correct stitch file, overlaid on the original artwork, does not line up. It is deliberately wider or longer in specific places, and the amount depends on the fabric, the stabilizer, the stitch type, and the machine.
Which means that checking a digitized file by overlaying the artwork and looking for exact registration tests the wrong thing entirely. A file that matches perfectly has probably not been compensated, and will pull inward on cloth. The reference that settles the question is a sewn sample on the actual fabric, and no comparison against the artwork substitutes for it — the general form of this problem being that a review is only as good as what it has to compare against: what review can and cannot catch.
The same point applies to automatic digitizing, and is the reason its failures are quiet. It produces a valid file that runs on a machine and looks broadly like the artwork. Whether the compensation, direction, and sequence are right is not something the file will report, and is not something anyone can see by looking at it on a screen.
Why “just make it smaller” means digitizing again
A request to run the same design at a different size sounds like scaling and is not, because the coordinates in a stitch file are absolute and the things that depend on absolute size do not scale with them.
Thread has a thickness that does not change. Stitches have practical minimum lengths that do not change. Detail that resolved at one size becomes a knot at a smaller one, densities that were correct become too tight or too loose, and small text that worked stops working. Scaling the geometry produces a file that will run and will not produce the same object.
The practical rule is that each size is its own job, and the design decisions that survive scaling are the ones made in the artwork rather than in the stitch file. Designs intended for several placements are cheaper when that is known at the start, because simplifications can be agreed once rather than improvised separately at each size.
Questions teams ask about digitizing
Which embroidery file format do we need? Whichever the machine running the job takes, which is a one-line answer from the supplier and rarely the thing worth discussing. The question with consequences is whether you are also receiving the editable digitizing source, because that determines whether future changes are edits or new jobs. Ask for both and store both.
Can we digitize from a photograph or a raster logo? Somebody can work from either, because digitizing is authoring rather than tracing and the input is a description of intent. A clean vector reduces the number of guesses required, which shows up as fewer surprises rather than as a different process. What matters more is whether anyone stated the things the artwork cannot carry.
Why does our embroidered logo look flat compared to the sample we liked? Fill direction is the usual answer. A design stitched in one uniform direction across every area loses the variation in reflected light that makes embroidery read as dimensional. It is a decision made during digitizing, it is reversible, and it is invisible in any file inspection.
Is automatic digitizing usable? It produces a file that runs, which is the difficulty rather than the recommendation. Direction, sequence, underlay, and compensation are decided by defaults rather than by someone looking at the specific garment and fabric, and none of that is visible until a sample exists. Where it is used, the sewn sample carries the whole burden of verification.
Can one digitized file work on both a cap and a jacket? Different fabrics distort differently and different placements constrain the machine differently, so compensation correct for one is wrong for the other. Treating each surface as its own job is the reliable approach. Reusing a file across surfaces is the version of this that produces a puckered result nobody can explain from the file.
What should we send along with the artwork? The fabric and garment it will be stitched on, the finished size, the placement, and anything about direction or emphasis you would otherwise leave to judgment. Those four close most of the gaps that an image cannot carry. Everything else is craft and belongs to whoever is doing it.
Where this leaves you
Digitizing is not a conversion with a format at the end of it. It is somebody deciding order, direction, underlay, and compensation — information no picture holds — and writing them as instructions for a machine. The practical consequences follow from that: get the editable source and not only the machine file, say what you care about instead of leaving direction to a stranger, treat each size and each fabric as its own job, and judge the result against a sewn sample rather than against the artwork it will deliberately fail to match.
Ask for the source, not just the file
When you next commission digitizing, ask for the editable digitizing source alongside the machine file, and check that you have actually received both. Send the garment, the fabric, the finished size, and the placement with the artwork, and say anything you have an opinion about regarding fill direction. Then approve on a sewn sample on the real fabric rather than on a screen preview or an overlay against your artwork.
Written by
