How the file is built

Layered SVG converter

A layered SVG here means one group per flat color, named and ordered from the back color to the front. The preset starts on Cricut cut. Change it if you are engraving, printing, or cutting vinyl.

Image converter

Drop a PNG, JPG, or WebP

You can also paste an image, or pick a file.

Nothing is uploaded. The picture stays in this browser tab.

Maker preset

Four flat colors, tight corners, and small specks removed. Each color is its own cut layer.

Drops islands smaller than this many pixels on a side. 0 keeps every speck.

Clears the color that dominates the edge, including a near-white edge. Spots of that color inside the design are cleared too.

Your layered preview shows up here. Hide a layer to leave it out of the download.

SVG keeps the curves. DXF writes polylines in millimetres at 96 pixels per inch, with curves flattened and Y flipped upright. PDF is one page, 1 pixel = 1 point. The zip is one SVG file per visible layer.

A Pro plan is planned for people who want unlimited layered files. It is not for sale, and there is no price yet. What that means

One group, one color

A lot of "image to SVG" files are a single soup of paths. That is enough if you only need a picture that scales. It is a poor cut file, because the machine software cannot tell red from blue, and you cannot hide the background without editing nodes by hand. This converter quantizes the image first, then traces it, then puts every path of a given fill into one SVG group. The group id and the Inkscape label are the same string, for example Layer 2 - #C2410C. Software that reads either of those will show the name.

The order is paint order. Layer 1 is drawn first. In a viewer that paints in file order, it sits at the back. The preview on this page uses that same SVG, so hiding a checkbox is the same operation as leaving that group out of the download. There is no second, secret version of the file.

What quantization is doing

A photograph can contain tens of thousands of colors, most of them a pixel of compression noise. A cutter cannot use that. Quantization picks a small palette, from 2 to 8 colors, and snaps every opaque pixel to the nearest one. If the picture already uses fewer flat colors than you asked for, those exact colors are kept. You do not get invented extra swatches to pad the count.

The palette is then handed to VTracer, which builds the paths. VTracer is the tracing engine. It is open source, MIT or Apache-2.0, and the copy in this site is the official WebAssembly build running in a worker. The page does not send pixels to a server to be traced somewhere else. If a trace fails, the status line on the converter is the error. There is no queue and no email when it finishes, because it finishes in the tab.

Holes, and what a layer is not

A layer is the visible region of that color. Where another color sits inside it, the path has a hole, using the nonzero fill rule. The red circle on a blue field becomes a red shape and a blue shape with a circular hole, not a blue rectangle with a red circle painted on top of a solid sheet. That is the right file for vinyl and for a laser that should not cut the same region twice. It is the wrong mental model if you wanted a solid backer sheet under every other color. This tool does not build that backer. You can draw one in Inkscape or Design Space from the outer contour if the job needs it.

Small holes and small islands are both at the mercy of the speckle filter. The number is a side length in pixels. The area cutoff is that number squared, both in the pixel cleanup and in VTracer's own speckle filter. A setting of 0 keeps every island, including the dust. A setting of 6 is aggressive and will eat small punctuation. Check the counters of letters after you move it.

Background removal

The option is deliberately simple. It looks at opaque pixels on the outer edge, finds the dominant color, and clears pixels within a tolerance of that color, anywhere in the image. If the edge is near white, the tolerance opens up so JPEG off-white goes with it. If the edge is transparent, there is no color to remove and the image is left alone. Enclosed spots of the edge color are removed too, which is what you want for the inside of an O on a white page, and what you do not want if that white is a highlight you meant to keep. Turn it off and convert again when the preview eats part of the design.

Picking an export

Download SVG when the next program is Design Space, Silhouette Studio, Inkscape, a browser, or Glowforge. The curves stay curves. Download DXF when the program wants polylines. Curves are flattened, units are millimetres at 96 pixels per inch, and the Y axis is flipped so the picture is upright in a CAD view. Inner closed polylines on a layer are the holes. Download PDF when a person needs to see or print the fills on one page. The page size is the artwork size in points, one SVG pixel per point. It is not a print-production PDF with bleed, crop marks, or a cut contour.

The zip is the same art split into files, one SVG per visible layer, using the layer name as the file name. Use it when a shop workflow wants each color as its own upload. Hiding a layer before you click the zip leaves that color out, the same as the other downloads.

A practical pass

Convert once with the preset. Hide the background layer if it is still there. Look at each remaining layer alone. If the edges crawl, lower smoothing. If dust remains, raise speckle size by one and convert again. If a color split into two layers that should be one, lower the color count. Then download. The 1200 pixel cap on the long side is there so the path count stays in a range a cutter can open. If you need a hairline from a huge scan, redraw that hairline as a vector. Tracing will not preserve it once the image has been scaled.

More specific notes: Cricut import, DXF for lasers and CNC, engraving artwork.

Questions

What does "back to front" mean here?

Layer 1 is painted first, so it sits at the back. Later layers paint on top. In the preview, hiding Layer 1 shows what was covering it. The download keeps the same order for the layers you leave visible.

Are the layers stacked sheets or separate pieces?

They are the visible region of each color, with holes cut out. They do not overlap to form a solid backer under the front shapes. If you need a solid sheet behind everything, that is a separate shape you add in your design software.

Why is a hole missing?

A hole is a subpath inside the color that surrounds it. If a counter was the same color as the background and background removal was off, it becomes its own layer instead of a hole. Turn removal on, or hide that layer. Very small holes can also be dropped by the speckle filter.

Can another site use the same converter code?

The image steps live in src/lib/pipeline and do not import the page. loadImage, removeBackground, quantize, vectorize, and the exporters are the whole API. The page is only one caller.

Which download should I use?

SVG if a cutter, browser, or Inkscape is the next stop. DXF if the machine software wants polylines. PDF if someone needs to look at or print the fills. The zip if each color must be its own file.