# From Image to Embroidery File: What Your Machine Actually Needs

> A DST or PES file encodes needle movements, not shapes — so no tool converts a JPG straight to stitches. Here is the honest pipeline, and where automated vectorization actually helps.

**Source:** https://vectosolve.com/blog/embroidery-file-from-image-what-machines-need
**Author:** Robin Monteiro (Founder of VectoSolve)
**Published:** 2026-07-15
**Last updated:** 2026-10-10
**Category:** Guide
**Reading time:** 11 min read
**Tags:** embroidery, digitizing, dst, pes, vectorization, file-formats

---

The pipeline below starts with turning your image into a clean vector — the [image to vector converter](/convert/image-to-vector) handles that first, automatable step.

## An embroidery file is machine code, not a picture

Search for "convert image to embroidery file" and most of the results imply the same workflow: upload a JPG, download a PES, stitch it out. That framing misunderstands what an embroidery file is, and it is responsible for a lot of the sagging, gap-riddled stitch-outs posted to machine-embroidery forums.

A stitch file (DST, PES, EXP, JEF, VP3) is not an image format. It is a sequence of machine instructions: move the hoop this far on X and Y, drop the needle, jump without stitching, trim, stop for a color change. Tajima's DST, the de facto commercial standard, records each stitch as a small X/Y displacement in 0.1 mm units and almost nothing else. The format reference maintained at EduTechWiki (University of Geneva) spells out just how spartan that is: a DST file is a fixed 512-byte header followed by stitch commands that are always exactly 3 bytes long, with a maximum stitch or jump length of 121 units — which, at 0.1 mm per unit, works out to about 12.1 mm of hoop travel per command. There is no "circle" in the file, no letter A, no fill color. Only movements.

That one fact explains everything else in this article. A JPG contains pixels. A stitch file contains choreography. Nothing in the pixels says whether a region should be a satin column or a tatami fill, how dense the rows should be, what stitches first, or how far the fabric will distort when thread pulls on it. Someone — or some software making the same judgment calls — has to decide. That step is called digitizing, and it is a separate craft from image conversion.

## The formats your machine actually reads

| Extension | Origin | Typically used by | What it stores |
|-----------|--------|-------------------|----------------|
| .dst | Tajima | Most commercial multi-needle machines | Movements and stops only; no thread colors |
| .pes | Brother | Brother and Baby Lock home machines | Stitches plus color and layout data |
| .exp | Melco | Melco, Bernina | Raw stitch data |
| .jef | Janome | Janome machines | Stitches plus color data |
| .vp3 | Husqvarna Viking | Viking, Pfaff | Stitches plus color data |
| .xxx | Singer | Singer machines | Stitches |

Two practical notes. First, DST stores no thread colors at all — "DST files do not contain colors," as the EduTechWiki format reference states verbatim: the file contains color-change stops, and the operator assigns actual thread cones at the machine. This is why a professional digitizer delivers a DST together with a color sequence sheet. Second, every format in this table encodes needle movements, not editable shapes — which is also why converting a stitch file back into artwork is lossy in the other direction.

Manufacturer documentation is worth reading before you buy or commission a design, because it is blunter than the marketing. Brother's official US support FAQ, which covers roughly 36 of its embroidery models including the PE535, SE2000, and PR1060W, lists exactly four design formats those machines will read — .pes, .dst, .phc, and .pen (the last from iBroidery) — and adds two constraints that derail more first stitch-outs than format confusion does. For the machines that FAQ covers, all designs must fit within a 7 × 5 inch field, and a design saved inside a folder on a USB stick cannot be retrieved by the machine at all: files have to sit in the root directory (according to Brother USA Support).

## The decisions a digitizer makes that no image contains

**Stitch type.** Machine embroidery has three primitives. Running stitch is a single line of stitches, used for fine lines and travel between regions. Satin is a zig-zag column with a distinctive sheen, used for borders and lettering, and it behaves well only in columns roughly 1 to 8 mm wide — wider than that, the long floats snag and sag. Fill (tatami) is offset rows of stitches for larger areas. Every region of your artwork must be assigned one of these.

**Density.** How tightly the rows pack. Too dense and the fabric puckers, the design goes stiff, and thread breaks climb. Too sparse and fabric shows through the fill.

**Underlay.** A hidden first pass of stitches that anchors the fabric to the stabilizer and lifts the visible stitching off the surface — essential on pile fabrics like fleece and terry. It is invisible in the finished piece and decisive for its quality.

**Pull compensation.** Stitches pull fabric inward along their direction, so a digitized circle stitches out as an ellipse unless the shapes are pre-widened against the pull. How much depends on the fabric: a stable cap front and a stretchy knit need different compensation.

**Pathing.** The order regions stitch, how the needle travels between them, and where trims happen. Bad pathing means visible travel stitches, more trims, and misregistration as the fabric shifts.

To see how far a stitch file drifts from the image that inspired it, run the arithmetic on a trivial case. A 50 × 50 mm filled square with fill rows 0.4 mm apart needs 125 rows; at a 4 mm stitch length, each 50 mm row takes about 13 needle penetrations. One flat square is already roughly 1,600 stitches — before underlay, borders, or a second color.

## The honest pipeline: raster → vector → stitches

Getting from an image to something a machine will sew is two distinct steps, and only one of them is reliably automatable today.

**Step 1: image to clean vector.** Tracing pixels into geometry — closed paths, flat solid colors, no gradients, no stray specks. This is exactly what automated vectorization is good at.

**Step 2: vector to stitches.** Digitizing, in dedicated embroidery software: assigning stitch types, setting densities, adding underlay, compensating for pull, planning the path. You can learn to do this yourself or send the art to a professional digitizer. Auto-digitizing features exist in embroidery software and behave best on precisely the kind of clean vector step 1 produces, but on complex art they still make stitch decisions a human has to correct.

Where VectoSolve sits in this pipeline, stated plainly: our [embroidery converter](/embroidery-converter) does step 1 only. It outputs vector files — SVG, EPS, DXF, and PNG — not stitch files. You cannot load our output into an embroidery machine. What you get is embroidery-ready artwork: a limited flat-color palette, closed filled shapes, speckles removed. That file then goes to your digitizer or into your digitizing software. Any tool that claims to skip step 2 entirely is either running an auto-digitizer (with the limitations above) or selling you a file your machine cannot use.

## Why clean vectors make digitizing cheaper

Digitizers quote per design, and the quote scales with how long the job will take. Send a low-resolution JPG with anti-aliased edges and a few hundred ambient colors, and the digitizer must first redraw your artwork before any digitizing starts — and that redrawing time is on your invoice. Send a clean vector with four flat colors and closed shapes, and the redraw disappears from the quote.

The going rates are public, and they are stitch-denominated — which is exactly why artwork complexity shows up on the bill. Absolute Digitizing, a commercial digitizing service, charges $1 per 1,000 stitches with a $10 minimum and a $40 cap per design, plus a flat $10 rate for a left-chest or cap design up to 4 × 4 inches, with standard turnaround within 12 hours (according to Absolute Digitizing's pricing page). EmbPunch, a USA-based service, charges $1.50 per 1,000 stitches with a $7.50 minimum covering up to 5,000 stitches and a maximum charge capped at $65, alongside flat tiers running from $11 to $49 by hoop size, with files typically ready in 2–8 hours (EmbPunch pricing page). Impact Digitizing quotes by garment placement, the way much of the trade does: $10 for a left-chest or cap logo up to 10,000 stitches, then $1 per 1,000 stitches beyond that, and $25 for a jacket back up to 50,000 stitches depending on complexity, with normal turnaround of 12–24 hours and urgent turnaround of 4–5 hours (Impact Digitizing pricing page). Two things stand out across those three price lists: the per-1,000-stitch convention holds at $1 to $1.50 across named vendors, and two of them independently land on $10 flat for a standard left-chest logo. Recall the arithmetic above — one flat 50 mm square is already about 1,600 stitches — and the incentive is clear: at stitch-denominated rates, every unnecessary filled region in your artwork is billed on every quote.

Clean art also cuts revision rounds. Ambiguity in a fuzzy image ("is this outline intentional? are these two blues supposed to be different threads?") turns into proof emails and days of back-and-forth. A vector with a fixed palette answers those questions before they are asked.

If you digitize your own designs, the same logic applies: auto-digitizing and manual punching both go faster when the input is unambiguous geometry rather than pixels.

The vector step is cheap to test. On VectoSolve you get one free watermarked preview a day with no signup; a clean conversion costs 0.20 credits, with plans starting at $7.99 a month for 25 conversions. The [embroidery vector converter](/c/embroidery-vector-converter) page covers the embroidery-specific settings in more detail.

## Fix these in the artwork before anyone digitizes it

- **Gradients.** Thread is a solid color. A gradient has to become either discrete bands of color or a manual thread-blending effect, which takes a digitizer real time. Flatten to a limited palette first; the embroidery preset does this automatically.
- **Text under 4–5 mm tall.** Thread has physical width. Below roughly 4–5 mm, letterforms fill in and turn into illegible blobs. Enlarge the text, simplify the typeface, or drop it (small taglines under logos are the usual casualty).
- **Photorealism.** Photographs depend on thousands of tones and soft transitions — the opposite of flat, stitchable regions. Stylize the image into a limited-palette illustration first, or accept that embroidery is the wrong medium for that image.
- **Hairline details.** Lines much under 1 mm cannot hold a satin column and fall back to a single running stitch, which reads far weaker than the artwork suggests. This limit is physics, not software preference: Melco's digitizing documentation for its DesignShop software sets minimum stitch dimensions of 10 points (1 mm) for satin width, 15 points for walk stitches, and 20 points for fills — because embroidery needles themselves are 7 to 8 points in diameter, any stitch under about 10 points will most likely need addressing before it sews cleanly (Melco service documentation).
- **Too many colors.** Commercial machines commonly carry 12 to 16 needles; home machines stitch one color at a time with manual thread changes. Every color is a stop and a trim. A typical left-chest logo works comfortably in two to six colors.

Once the palette is fixed, map it to real thread: the free [thread color mapper](/thread-color-mapper) matches your artwork's colors to Madeira and Isacord thread numbers, so the color sheet you hand over references cones that actually exist.

## What to send a digitizer

A short checklist that saves a proofing round: the vector file (SVG or EPS), the finished size ("90 mm wide, left chest"), the garment and fabric type (density, underlay, and compensation all depend on it), the thread color numbers, and the placement. Fabric matters more than most first-timers expect — the same file digitized for a twill cap will not sew well on a knit polo.

## FAQ

**Can software convert a JPG straight to a PES file with no human involved?**
Auto-digitizing features exist and keep improving. On simple, flat-color artwork at a sensible size they can produce stitchable results. On anything complex, the stitch decisions — density, underlay, pull compensation, pathing — still need a human eye. The step that is reliably automatable today is raster to vector, not vector to stitches.

**Does VectoSolve output DST or PES files?**
No. VectoSolve outputs SVG, EPS, DXF, and PNG: artwork files that are the input to digitizing, not the output of it. We state this plainly because tools that blur that line cost embroiderers money and time.

**Which stitch format should I ask for?**
Whatever your machine's manual lists. DST is the safe default for commercial multi-needle machines; PES for Brother and Baby Lock home machines; JEF for Janome; VP3 for Viking and Pfaff. If your digitizer delivers DST, expect a separate color sheet, because DST stores no thread colors.

**Why does my machine reject my SVG?**
Because an SVG describes shapes and an embroidery machine executes needle movements. Until digitizing assigns stitches to those shapes, there is nothing in the file the machine can act on. That is not a defect in your file — it is the difference between artwork and machine instructions.

## Sources

- [Absolute Digitizing pricing page](https://absolutedigitizing.com/embroidery-digitizing-pricing/) — per-1,000-stitch digitizing rates ($1/1,000, $10 minimum, $40 cap), the flat $10 left-chest/cap rate up to 4 × 4 inches, and 12-hour standard turnaround.
- [EmbPunch embroidery digitizing pricing page](https://www.embpunch.com/embroidery-digitizing) — $1.50 per 1,000 stitches, $7.50 minimum covering up to 5,000 stitches, $65 cap, flat tiers from $11 to $49, and 2–8 hour delivery.
- [Impact Digitizing pricing page](https://impactdigitizing.com/pricing/) — placement-based flat pricing ($10 left chest/cap up to 10,000 stitches, $25 jacket back up to 50,000 stitches) with 12–24 hour normal and 4–5 hour urgent turnaround.
- [Brother USA Support — "Design format types that the machine will read"](https://help.brother-usa.com/app/answers/detail/a_id/123343/~/design-format-types-that-the-machine-will-read) — the four readable formats (.pes, .dst, .phc, .pen), the 7 × 5 inch field limit for the models covered, and the USB root-directory requirement.
- [EduTechWiki (University of Geneva), "Embroidery format DST"](https://edutechwiki.unige.ch/en/Embroidery_format_DST) — DST file structure: fixed 512-byte header, 3-byte stitch commands, absence of color data, and the 121-unit maximum stitch/jump length.
- [Melco service documentation — "Stitches Too Small"](https://www.melco-service.com/source1/Stitches_Too_Small.htm) — minimum stitch dimensions in DesignShop guidance (satin ≥ 10 points/1 mm, walk ≥ 15 points, fill ≥ 20 points) and the 7–8 point needle diameter.

---

*VectoSolve — AI-powered SVG platform: image-to-vector conversion, AI animation, brand identity generation, and more. https://vectosolve.com*

---

*Maintained by VectoSolve. When citing, link to https://vectosolve.com/blog/embroidery-file-from-image-what-machines-need.*

*VectoSolve — AI-powered SVG platform: image-to-vector conversion, AI animation, brand identity generation, and more. https://vectosolve.com*