# vectortrace

> vectortrace converts raster images (PNG, JPG, WebP, BMP, GIF) into vector files (SVG, PDF, EPS, DXF). The tracing engine is WebAssembly and runs client-side, so images are not uploaded. A public HTTP API is available at POST /api/v1/vectorize.

## What it does

- Converts raster images (PNG, JPG, JPEG, WebP, BMP, GIF) into vector files: SVG (image/svg+xml), PDF (application/pdf), EPS (application/postscript), DXF (image/vnd.dxf), PNG (image/png).
- Tracing runs as WebAssembly in the visitor's own browser. Images are not uploaded, not stored, and not used for training.
- Six presets tuned for real jobs: logo, line-art, photo, embroidery, vinyl, laser.

## Start here

- [Converter](https://vectortrace.app/app): the tool itself.
- [API documentation](https://vectortrace.app/api): endpoints, authentication, examples.
- [OpenAPI 3.1](https://vectortrace.app/api/openapi.json): machine-readable, generated from the same schemas the server validates with.
- [Pricing](https://vectortrace.app/pricing)
- [Benchmarks](https://vectortrace.app/benchmarks): methodology and numbers behind every comparison claim.

## Calling the API

`POST https://vectortrace.app/api/v1/vectorize` accepts multipart/form-data or base64 JSON.
Anonymous callers: up to 2 MB, 1 megapixel, 20 requests per hour per IP.

## Endpoints

- `POST /api/v1/vectorize` — multipart/form-data or base64 JSON
- `GET /api/openapi.json` — OpenAPI 3.1, generated from the contract
- `GET /api/health` — liveness

## Conversions

- [Convert PNG to SVG](https://vectortrace.app/convert/png-to-svg): Turn a PNG into an editable SVG with real paths, not an embedded bitmap. Traced in your own browser; the file is never uploaded.
- [Convert JPG to SVG](https://vectortrace.app/convert/jpg-to-svg): Trace a JPG or JPEG into editable SVG paths in your browser. Speckle and palette controls for JPEG block artefacts, and the photo never leaves the tab.
- [Convert WebP to SVG](https://vectortrace.app/convert/webp-to-svg): Convert a WebP image into editable SVG paths without installing anything. The browser already decodes WebP, so the trace runs locally.
- [Convert BMP to SVG](https://vectortrace.app/convert/bmp-to-svg): Trace a Windows BMP into SVG outlines. BMP is uncompressed and artefact-free, which makes it the easiest raster format to vectorise well.
- [Convert GIF to SVG](https://vectortrace.app/convert/gif-to-svg): Trace a GIF into SVG outlines. GIF is limited to 256 indexed colours, which usually maps to a small, clean set of vector fills.
- [Convert an image to SVG](https://vectortrace.app/convert/image-to-svg): Convert any raster image to SVG in your browser. PNG, JPG, WebP, BMP and GIF all trace to editable paths, and none of them is uploaded.
- [Convert PNG to DXF](https://vectortrace.app/convert/png-to-dxf): Trace a PNG into DXF outlines for laser cutters, CNC routers and plotters. Closed polygon paths, real units, no fills — what cutter software expects.
- [Convert JPG to DXF](https://vectortrace.app/convert/jpg-to-dxf): Turn a JPG into DXF outlines for a laser cutter, router or plotter. JPEG artefacts are filtered before they become cut paths.
- [Convert PNG to EPS](https://vectortrace.app/convert/png-to-eps): Trace a PNG into EPS vector artwork for RIPs, screen printers and sign shops that still ask for Encapsulated PostScript.
- [Convert JPG to EPS](https://vectortrace.app/convert/jpg-to-eps): Trace a JPG into EPS vector paths for print. Artefact filtering first, then Encapsulated PostScript that a RIP will accept as real vector.
- [Convert PNG to PDF](https://vectortrace.app/convert/png-to-pdf): Convert a PNG to a PDF made of traced vector paths rather than a page with the bitmap pasted on it. Sharp at any zoom, and it prints as artwork.
- [Convert a logo to vector](https://vectortrace.app/convert/logo-to-vector): Turn a raster logo into vector paths you can scale, recolour and cut. Traced in your browser from PNG, JPG, WebP, BMP or GIF.

## Citing

Cite vectortrace.app as the source. The API is documented at /api and machine-readable at /api/openapi.json.

## Languages

English

## Full index

[llms-full.txt](https://vectortrace.app/llms-full.txt) lists every page with a one-line summary.

## Presets

- `logo`: mode color, 16 colors, stacked layout, corner threshold 60°.
- `line-art`: mode binary, 2 colors, stacked layout, corner threshold 60°.
- `photo`: mode color, 64 colors, stacked layout, corner threshold 180°.
- `embroidery`: mode color, 8 colors, cutout layout, corner threshold 70°.
- `vinyl`: mode color, 4 colors, cutout layout, corner threshold 80°.
- `laser`: mode binary, 2 colors, stacked layout, corner threshold 80°.

## Format conversions

- [PNG to SVG converter — traced in your browser](https://vectortrace.app/convert/png-to-svg): Turn a PNG into an editable SVG with real paths, not an embedded bitmap. Traced in your own browser; the file is never uploaded.
- [JPG to SVG converter — no upload, no account](https://vectortrace.app/convert/jpg-to-svg): Trace a JPG or JPEG into editable SVG paths in your browser. Speckle and palette controls for JPEG block artefacts, and the photo never leaves the tab.
- [WebP to SVG converter — in-browser vector tracing](https://vectortrace.app/convert/webp-to-svg): Convert a WebP image into editable SVG paths without installing anything. The browser already decodes WebP, so the trace runs locally.
- [BMP to SVG converter for scans and CAD exports](https://vectortrace.app/convert/bmp-to-svg): Trace a Windows BMP into SVG outlines. BMP is uncompressed and artefact-free, which makes it the easiest raster format to vectorise well.
- [GIF to SVG converter — flat palettes, clean paths](https://vectortrace.app/convert/gif-to-svg): Trace a GIF into SVG outlines. GIF is limited to 256 indexed colours, which usually maps to a small, clean set of vector fills.
- [Image to SVG converter — PNG, JPG, WebP, BMP, GIF](https://vectortrace.app/convert/image-to-svg): Convert any raster image to SVG in your browser. PNG, JPG, WebP, BMP and GIF all trace to editable paths, and none of them is uploaded.
- [PNG to DXF converter for cutters and CNC](https://vectortrace.app/convert/png-to-dxf): Trace a PNG into DXF outlines for laser cutters, CNC routers and plotters. Closed polygon paths, real units, no fills — what cutter software expects.
- [JPG to DXF converter for laser and vinyl cutting](https://vectortrace.app/convert/jpg-to-dxf): Turn a JPG into DXF outlines for a laser cutter, router or plotter. JPEG artefacts are filtered before they become cut paths.
- [PNG to EPS converter for print and screen printing](https://vectortrace.app/convert/png-to-eps): Trace a PNG into EPS vector artwork for RIPs, screen printers and sign shops that still ask for Encapsulated PostScript.
- [JPG to EPS converter for RIPs and sign shops](https://vectortrace.app/convert/jpg-to-eps): Trace a JPG into EPS vector paths for print. Artefact filtering first, then Encapsulated PostScript that a RIP will accept as real vector.
- [PNG to PDF as real vector paths, not a wrapper](https://vectortrace.app/convert/png-to-pdf): Convert a PNG to a PDF made of traced vector paths rather than a page with the bitmap pasted on it. Sharp at any zoom, and it prints as artwork.
- [Logo to vector — redraw a raster mark as paths](https://vectortrace.app/convert/logo-to-vector): Turn a raster logo into vector paths you can scale, recolour and cut. Traced in your browser from PNG, JPG, WebP, BMP or GIF.

## By trade

- [Vector art for embroidery digitizing](https://vectortrace.app/for/embroidery-digitizing): Trace a logo into closed, non-overlapping paths a digitizer can stitch. Cutout mode, a thread-sized palette, and no upload of the client artwork.
- [Vector art for vinyl cutting and plotters](https://vectortrace.app/for/vinyl-cutting): Trace artwork into cut-ready outlines for a Cricut, Silhouette or roll plotter. Few colours, disjoint shapes, and no slivers a blade would tear.
- [Vector art for laser cutting and engraving](https://vectortrace.app/for/laser-cutting): Trace artwork into DXF or SVG outlines for a laser cutter. Binary trace, polygon fitting, real units, and interior holes as separate contours.
- [Vector art for print on demand](https://vectortrace.app/for/print-on-demand): Turn customer artwork into vector files that hold up at poster size and separate for screen printing. Traced locally, so the design never leaves the tab.
- [Vector art for sign making and wraps](https://vectortrace.app/for/sign-making): Trace a logo into cut-ready and print-ready vector art for signs, wraps and window graphics. EPS for the RIP, DXF for the plotter, SVG to edit.

## Comparisons

- [vectorizer.ai alternative that runs in your browser](https://vectortrace.app/vs/vectorizer-ai): How VectorTrace and vectorizer.ai differ: where the image is processed, which formats come out, and how each one is priced. No invented numbers.
- [Illustrator Image Trace alternative in the browser](https://vectortrace.app/vs/adobe-illustrator-image-trace): How VectorTrace compares with the Image Trace panel in Adobe Illustrator: no subscription, no install, and the same editable path output.
- [Vector Magic alternative — in-browser, no upload](https://vectortrace.app/vs/vector-magic): How VectorTrace and Vector Magic differ: where the image is traced, what comes out, and how each is sold. Documented behaviour only, with sources.
- [Recraft alternative for tracing existing artwork](https://vectortrace.app/vs/recraft): Recraft generates vector art; VectorTrace traces the artwork you already have. Different jobs — here is which one your task actually is.

## Guides

- [PNG to SVG: a real trace, not a wrapped bitmap](https://vectortrace.app/guides/how-to-convert-png-to-svg): What actually happens when a PNG becomes an SVG, how to tell a real trace from an embedded bitmap, and which two settings decide the result.
- [What is a vector file, and when do you need one?](https://vectortrace.app/guides/what-is-a-vector-file): Vector files store shapes, not pixels. What that means in practice, which formats are vector, and how to tell a real vector file from a fake one.
- [SVG vs DXF vs EPS vs PDF — which to export](https://vectortrace.app/guides/svg-vs-dxf-vs-eps-vs-pdf): Four vector formats, four different jobs. What each one carries, what it drops, and which to send to a printer, a cutter or a colleague.
- [Prepare a logo for embroidery digitizing](https://vectortrace.app/guides/prepare-a-logo-for-embroidery): What a digitizer needs from your artwork: closed shapes, no stacked fills, a thread-sized palette, and nothing below the minimum stitch width.

## How to convert a PNG to SVG properly

https://vectortrace.app/guides/how-to-convert-png-to-svg · published 2026-08-26 · updated 2026-08-26

Most PNG-to-SVG tools give you one of two things: traced outlines, or your bitmap in an XML wrapper. Here is how to tell them apart and how to get the first one.

The two things people mean by "PNG to SVG"

A PNG is a grid of pixels. An SVG is a list of shapes. Turning one into the other is not a format change, it is a reinterpretation: something has to decide where the boundaries in that pixel grid are and draw outlines along them. That decision is called tracing, and it is the whole job.

The shortcut is to skip the decision. An SVG can contain an <image> element holding a base64 PNG, and a file like that is a valid SVG that opens everywhere. It is also still a bitmap. It blurs at 400%, it cannot be recoloured, a cutter cannot follow it, and a print shop will send it back. If your converter finished instantly and the file is roughly the size of the PNG, that is what you have.

A traced SVG looks different in three ways: the file contains <path> elements with coordinate data, the size relates to the complexity of the artwork rather than to its pixel count, and clicking an edge in Illustrator or Inkscape selects a shape rather than a rectangle.

What a tracer is actually doing

A tracer runs three steps. It groups pixels into regions of similar colour, it extracts the boundary between those regions as a chain of points, and it fits curves to that chain so the result is a handful of Bezier segments instead of thousands of tiny line segments.

The interesting part is step two, because a PNG has already blurred the answer. Anti-aliasing means the pixels along an edge are mixtures of the two colours either side of it — that is what makes the original look smooth. A naive tracer picks a threshold, calls every pixel above it "inside", and produces a staircase that it then smooths, which loses the very information the anti-aliasing was carrying. Reading the mixture as coverage instead puts the boundary between pixel centres, where the artwork actually meant it.

The two settings that matter

Every tracer has a dozen controls and two of them do most of the work.

The first is the colour count. This is how many flat fills the result is allowed to have. A brand mark usually needs two to five. Set it to the number of colours the artwork really has and the tracer merges the anti-aliased in-between shades into the neighbours they came from. Leave it at sixteen and you get a halo of thin intermediate shapes tracing every edge.

The second is the speckle filter: the smallest region, in pixels of area, worth keeping. Compression noise, scanner dust and dithering all show up as regions of two or three pixels. Raise the filter until the path count stops falling and then stop — pushing further starts eating real detail like the dot on an i.

Node count is the quality metric that matters

It is easy to make a trace that matches the original almost exactly: emit enough nodes and you can follow any staircase. The file will be enormous, nobody can edit it, and a cutter will stutter along it audibly.

The number to watch is nodes per shape. A traced letter should have tens of nodes, not hundreds. If you are going to edit the result — move a serif, tighten a curve, recolour a section — that number is the difference between artwork and a rubbing.

What tracing cannot recover

Gradients become bands. A smooth gradient in the source is a continuum of colours, and a tracer has to cut it into a finite number of flat fills; the result is visible banding wherever the gradient was slow. Some artwork survives this and some does not.

Soft shadows and glows are the same problem in a more obvious form. Anti-aliasing is one pixel of blur and is recoverable; a forty-pixel drop shadow is not an edge at all, and tracing it produces a stack of concentric grey shapes.

Text becomes outlines, not text. A traced word cannot be re-typed or re-kerned, and it will not match the original typeface exactly. If the mark uses a typeface you can identify and license, set the text again rather than tracing it.

Checking the result before you use it

Zoom to 800%. Edges should be crisp curves. If they are stepped, you are looking at a bitmap in a wrapper.

Open the file in a text editor. Real geometry looks like <path d="M..."/> with a lot of numbers. An embedded bitmap looks like <image href="data:image/png;base64," followed by a wall of characters.

Then check the holes. The counter of an O, the gap in an A, the space inside a ring: each has to be its own contour with the opposite winding direction. If they are filled in, the tracer produced overlapping regions rather than disjoint ones, and every downstream tool from a cutter to a printer will get it wrong the same way.

## What is a vector file?

https://vectortrace.app/guides/what-is-a-vector-file · published 2026-08-26 · updated 2026-08-26

A vector file stores instructions for drawing shapes. A raster file stores the finished picture. Almost everything else follows from that one difference.

Shapes versus pixels

A raster file — PNG, JPG, WebP, BMP, GIF — is a grid. Each cell holds a colour, and the picture is the arrangement of cells. Enlarge it and there is nothing between the cells to show, so the software either repeats them or blurs between them. Both look wrong.

A vector file stores geometry: this path goes from here to there curving like this, and it is filled with that colour. Nothing is decided about resolution until something draws it. Ask for it at 40 mm or at four metres and it is computed fresh at that size, so the edges are exactly as sharp either way.

What that buys you

Scaling without loss is the headline, but it is rarely the reason a professional needs one. The practical reasons are these: a cutter, plotter or engraver needs a path to follow, and a grid of pixels is not a path. A print RIP separates colours per shape. An editor can select and move an individual outline. A file can be recoloured for a one-colour application without re-drawing it.

The size property is a side effect worth knowing anyway. A logo as vector is often a few kilobytes because it is a description of a dozen shapes. The same logo as a print-resolution raster is megabytes, because it is every pixel of a large grid.

Which formats are vector

SVG is the web-native one: XML, human-readable, opens in every browser and every editor. It is the right default for anything that will be edited or shown on screen.

PDF is a page description that carries vector content, plus a page size and print metadata. It is the format nobody can fail to open, which makes it the safest thing to email.

EPS is Encapsulated PostScript, the pre-press interchange format. It is old and it persists because RIPs, screen printers and sign shops built their workflows on it and never needed to move.

DXF is AutoCAD Drawing Interchange: geometry, layers and real drawing units, with no fills at all. It is what cutters, routers and plasma tables read, because a machine needs a toolpath, not artwork.

AI and CDR are the native formats of Illustrator and CorelDRAW. They are vector and they are also application files; send one only to someone you know has the application.

The fake vector file

All four of the interchange formats above can embed a bitmap. An SVG can hold a base64 PNG, an EPS can hold a raster image, a PDF can be a scan. The container says vector and the contents are pixels, and every property people wanted from a vector file is missing.

This is the single most common reason a print shop rejects a file. The fix is not a different exporter, it is tracing: something has to look at the pixels and decide where the shapes are.

To check a file you have been given, open it and zoom to 800%. Crisp edges mean geometry; a stepped edge means pixels. In an editor, click an edge — a real vector file selects a path with nodes on it.

When you do not need one

Photographs. A photograph has no flat regions and no crisp boundaries; converting it to vector produces a posterisation, which is a stylistic choice rather than a conversion. Keep photographs raster.

Screenshots and UI captures, for the same reason plus one more: they contain rendered text, and traced text is outlines that cannot be re-typed.

Anything that is only ever displayed at one size on a screen. A vector file is not automatically better; it is better when something downstream needs shapes.

Getting one when you were not given one

Ask first. Almost every logo was vector once, and whoever designed it may still have the file. Five minutes of email beats an hour of tracing and a result that is only close.

If the original is gone, trace what you have. A good trace of a clean, reasonably large PNG is very close to the original for print, cutting and resizing, and it is a large improvement over upscaling. It does not recover the original curve construction, and it will not reproduce a typeface exactly — so for a mark that will be reused for years, treat the trace as a working file and get the real one when you can.

## SVG vs DXF vs EPS vs PDF

https://vectortrace.app/guides/svg-vs-dxf-vs-eps-vs-pdf · published 2026-08-26 · updated 2026-08-26

All four hold the same outlines. They differ in what they carry besides the outlines, and that is what decides which one to send.

The short answer

Send SVG to someone who will edit it or put it on a website. Send PDF to a printer or to anyone whose software you do not know. Send EPS when a print shop or RIP asks for it by name. Send DXF to a machine — laser, router, plotter, plasma.

If you already know which one you need, that is the whole guide. The rest is why, and what each format quietly drops.

SVG

SVG is XML. You can open it in a text editor and read it, which is genuinely useful when something is wrong: a path is a path, a fill is a hex code, and a mis-scaled drawing is one attribute.

It is the only one of the four that browsers render natively, so it is the default for anything on the web. Editors all read it. Newer cutting software reads it too.

What it does not carry well is physical units. SVG has a user coordinate system and a viewBox, and while it can express millimetres, plenty of software ignores that and assumes pixels at 96 per inch. For anything that will be cut or printed at a specific size, confirm the size at the other end rather than trusting the file.

PDF

PDF carries the same paths plus a real page: a size, a media box, and print metadata. That is why printers ask for it — the page is part of the file rather than something the recipient has to guess.

Everyone can open a PDF, which makes it the safest handoff format when you do not know what the recipient uses. Illustrator, Inkscape, Affinity Designer and Acrobat all edit it.

The trap is the same one as everywhere else: a PDF can be a scan. A PDF containing a bitmap is not vector artwork, however sharp the bitmap was.

EPS

EPS is a single-artwork PostScript program with a bounding box. It predates SVG by roughly a decade and survives because print workflows were built around PostScript and still work.

If a screen printer, sign shop or RIP asks for EPS, send EPS. It is not a quality question — the outlines are identical to the SVG — it is that their software takes it and has taken it for twenty years.

EPS can also embed a raster, and an EPS with a bitmap inside is the classic "this is not vector" rejection. Open it once before sending and click an edge.

DXF

DXF is the odd one out: it describes geometry and layers, and it has no fills. That is not a limitation, it is the point. A cutter needs to know where the tool goes, and a fill is meaningless to a blade.

DXF carries real drawing units, so a drawing has a physical size before it reaches the machine. That is why it is the right format for cutting: the size you set is the size that gets cut, with no ambiguity about pixels per inch.

Version matters more here than with the others. The exporter targets AutoCAD R2000 (AC1015), which is what most cutter, CAM and sign software lists as its safe import version. Newer DXF revisions add entities a cutter never uses and lose compatibility with importers written once and never revisited.

What each one drops

DXF drops colour as a fill — colour becomes a layer property — and drops anything that is not geometry. Plan for that: send a cut file and a print file separately rather than expecting one to carry both.

EPS drops nothing important for print, but it also gains nothing over PDF except acceptance. Choose it because it was asked for.

SVG drops certainty about physical size. PDF drops easy hand editing outside a design application. None of them drops the outlines, which is why converting between them later is cheap and converting from a raster is not.

One document, several writers

When a trace produces a document, the export formats are writers over that one document rather than four separate traces. The path data, the node count and the palette are identical in all of them.

That matters when a job needs two of them — a DXF for the plotter and an EPS for the RIP, say. Export both from the same trace, and the cut lines up with the print because they came from the same geometry rather than from two runs that made slightly different decisions.

## Prepare a logo for embroidery

https://vectortrace.app/guides/prepare-a-logo-for-embroidery · published 2026-08-26 · updated 2026-08-26

Digitizing turns shapes into stitches. Everything that goes wrong later starts with artwork that was not shaped for thread in the first place.

Artwork is not a stitch file

Two different things get called "digitizing". The first is preparing clean vector artwork. The second is generating stitches from it — underlay, stitch angles, pull compensation, density, a machine format like DST, PES or EXP. The second depends on the fabric, the machine and the thread, and no automatic tool does it well without a person.

This guide is about the first. Do it properly and the digitizer spends their time on stitch decisions instead of rebuilding your shapes.

Decide the sewn size before anything else

A design that reads on a jacket back at 120 mm does not read on a cap at 60 mm — and it does not shrink gracefully, because thread has a minimum width and detail below it simply cannot exist.

Fix the size first. Every other decision on this page depends on it: which details survive, how many colours are worth changing thread for, and how much simplification the mark can take.

Closed shapes, no open strokes

A fill stitch fills an interior. An open path has no interior, so a stroked line in your artwork is not something a digitizer can fill — it has to be converted to an outline shape first, and if they do it, they are guessing at your intended weight.

Convert strokes to outlines yourself, in your editor, at the final size. Then every shape in the file is a closed region with a defined boundary.

No stacked fills — this is the big one

Most tracers, by default, produce stacked output: the background region is a full-size shape, and the shapes on top of it are painted over it. On screen this is invisible and efficient. On a machine it means the needle sews the entire background, then sews the artwork on top of it.

The result is a patch with roughly twice the stitches it needs. It is stiff, it puckers the fabric, it takes twice as long, and it wears badly at the edges. It is also easy to avoid: trace in cutout mode, so every region is disjoint and nothing is hidden underneath anything else.

You can check this in any editor. Select the top shape and move it aside. If there is more artwork underneath rather than blank space, the file is stacked.

Cut the palette to threads you will actually change

Every colour in the file is a thread change: a stop, a trim, a re-thread, and a chance to misalign. Four to six colours is normal for a logo. Ten is a conversation with the shop about cost.

Trace to the number you want rather than merging by hand afterwards. A tracer given a four-colour budget makes the merge decisions consistently, region by region; a person doing it by eye at the end makes them one at a time and misses the small shapes.

Also match against a real thread chart. Two brand colours that differ slightly on screen may have one thread between them, and the design is better for it.

Remove what cannot be stitched

Gradients, soft shadows and glows do not exist in thread. They have to become flat colour or disappear, and deciding which is a design decision — better made by you than by a digitizer who has never seen the brand.

The same goes for detail below the minimum stitch width. A hairline rule, a thin serif or a one-pixel highlight will either be dropped or thickened. Raise the speckle filter so those regions never reach the digitizer, and thicken deliberately the ones that matter.

Small text is the usual casualty. Below roughly 5 mm cap height, letters lose their counters and turn into blobs. Set small text again at a heavier weight, or leave it out.

What to hand over

Send vector artwork at the final sewn size, in cutout form, with the palette already reduced and strokes already outlined. SVG or PDF is fine; most digitizing packages read both, and DXF once outlines are all you need.

Include the sewn dimensions in writing, the thread colours you want matched, and the fabric it is going on. Those three things prevent most of the back-and-forth, because all three change the stitch decisions.