Rasterless

Turn a bitmap into real vector paths

Three meanings of bitmap, one answer: curves fitted to the pixel grid.

Convert and preview without an account. Downloading is free too — it just needs a Google sign-in. No card, no subscription.

Examples
SOURCE VECTOR A round badge with soft tonal shading, traced into flat vector colour bands — source image
A round badge with soft tonal shading, traced into flat vector colour bands — vector result

01 · 5 colours · 193 nodes

How it works

  1. Check the file is a PNG, JPEG, WEBP, BMP, GIF or AVIF — a .bmp goes in as it is; a TIFF needs re-saving as PNG first.
  2. Drop it on the converter and set a colour count: 12 to start, or 2 for a single-colour silhouette.
  3. Choose your outputs — SVG for editing, PDF or EPS for print, DXF for cutting; pick several and you also get a ZIP.
  4. Convert and look at the preview. If edges look chunky or detail is missing, change the colour count and run it again.
  5. Sign in with Google to download the vector files; they are free and carry no watermark.

"Bitmap" is one word doing three jobs, and which one you have in mind changes what you should do next. In Inkscape and GIMP it simply means an image made of pixels — that is the sense behind Path > Trace Bitmap, which is where a lot of people first meet the idea of turning a picture into paths. In older graphics work, and in typeface work, it means one bit per pixel: black where the bit is set, white where it is not.

On Windows it is also a file extension, the uncompressed pixel format that still falls out of scanner utilities, CAD viewers and screen-capture tools.

That last meaning needs no extra step: a file whose name ends in .bmp goes straight in. Its first bytes are checked to confirm it really is a BMP, and only the BMP decoder is allowed to open it. TIFF is the exception — it stays out on purpose, to keep the number of image decoders the server has to trust small, so re-save a TIFF as PNG first. PNG stores the identical pixel grid losslessly, so that round trip throws nothing away. What the uploader takes is PNG, JPEG, WEBP, BMP, GIF and AVIF.

For the other two meanings the work on this side is identical: read the pixel grid, decide where the real colour boundaries sit, and write those boundaries out as curves. SVG, PDF, EPS and DXF all come from that same traced geometry.

What actually happens

What actually happens
  • Raster 6×6 · 8 px · PNG 77 B
  • 6 nodes · 1 path
  • SVG 91 B · vector, not pixels

What the bitmap's pixel grid does to the trace

A one-bit bitmap has no edge ramp to measure

A one-bit image has no blended pixels along its edges, so the anti-aliasing model has nothing to read and the boundary falls on the crack between black and white. Diagonals arrive as a staircase of pixel steps that the curve fitter smooths within tolerance. Resolution matters more here than for anti-aliased artwork: a larger scan means smaller steps and a straighter line.

Pixel art is smoothed unless each bitmap pixel becomes a block

The fitter keeps the corners it can detect and smooths the rest, so a sprite traced at its native size comes back with straightened diagonals and rounded curves, the pixel look gone. If the squares are the point, scale the sprite up first with nearest-neighbour resampling, so each pixel becomes a block the fitter reads as four corners. A 64 px sprite at 16× stays under 1.5 megapixels.

Dust on a scanned bitmap becomes closed paths

At 2 colours every dark speck on a scanned stamp is a region, and every region becomes a closed path, so dust can come through as a scatter of tiny shapes around the mark. The colour count is not the control for that, because the specks are the same black as the artwork. Clean the scan first, or delete the strays in your editor afterwards.

Choosing the colour count

A pixel grid has one size baked into it: enlarge it and the edges go soft or the squares start showing. A path carries no such ceiling, because the renderer draws the curve fresh at whatever scale you ask for. That is the whole argument for tracing, and it is why an old icon, a scanned stamp or a one-bit logo that never existed above screen size can still end up on a shirt, a sign or a cutting mat.

There is a fair suspicion attached to tools like this, and it deserves a direct answer: did it really redraw the artwork, or did it park the original pixels inside a vector container and call it done? You can settle that in a text editor. The SVG here holds path elements and nothing else — no image element, no base64 blob, no PNG smuggled inside. Every edge is a curve fitted to a colour boundary the tracer measured in your pixels, which is why the file stays small and why recolouring a shape in an editor actually recolours it.

That matters most to people who already had the word "bitmap" in their vocabulary before they searched for it. If you have been running Trace Bitmap locally and want the result to drop into a drawing you are already editing, ask for SVG grouped by colour: each colour arrives in its own group, so selecting every blue shape is one click rather than a rubber-band and a lot of shift-clicking.

What it will not do is invent detail that was never in the grid. It does not read lettering as text, it does not remove a background beyond what the alpha channel already marks transparent, and it takes one image per conversion.

Settings we suggest

The preset on this page is 12 colours, Cut-outs and filled shapes, which suits the flat artwork that "bitmap" usually turns out to mean. Colour count is the control that moves the result most, and it runs from 2 to 64. Two is the setting for a genuine one-bit bitmap, or for anything you want as one clean silhouette. A dozen or so holds a flat logo or icon together. Go high only for shaded sources, and expect bands rather than a smooth gradient — the sample on this page is a badge with soft tones, and each tone becomes its own region.

Shape stacking decides how those regions relate. Cut-outs removes the area covered by the shapes above it, so every region is a separate closed contour and nothing overlaps; that is what keeps a cutting path sane. Stacked leaves the lower shapes whole and lets the upper ones sit on them.

Draw style is fill_shapes here, so each colour area becomes a filled path; the two stroke styles draw the artwork as lines instead. Group by colour changes how the SVG is grouped, not the geometry.

Choose formats before converting — picking more than one adds a ZIP with everything in it. DXF is R2010 (AC1024) and carries outlines only: LWPOLYLINE and SPLINE entities, one layer per colour, with millimetres declared in the header so the machine opens the part at size instead of guessing the unit. The PNG output is not a vector; it is the traced result rendered back to pixels on a transparent background, which is handy as a cleaned-up copy. For a physical width and height there is a millimetre field, and it reaches all four vector formats — the SVG and DXF geometry, and the page box of the PDF and EPS — so a size you ask for here is the size the next tool sees.

Questions

What do I do with a file that ends in .bmp?

Upload it as it is. The first bytes are checked to confirm it really is a BMP before anything decodes it, and the trace comes out the same as for the identical pixels saved as PNG.

Does the SVG still contain the original pixels somewhere?

No. Open it in a text editor and you will find path elements and nothing else: no image element, no embedded picture data. The shapes were fitted to colour boundaries measured in your bitmap, and the bitmap itself is not carried along.

Is there a learned model doing the tracing?

Partly. The tracing is geometric (it segments the image into colour regions, uses an explicit anti-aliasing model to work out where an edge really sits between two pixels, then fits curves to the result), and it is combined with machine learning (AI) that runs on our own servers. Your image is not used for anything else.

How large can the bitmap be, and how long does it take?

Up to 25 MB and 40 megapixels. Images above 1.5 megapixels are downscaled before tracing, so the output's pixel dimensions are the downscaled ones — use the millimetre field when you need an exact physical size. Simple logos usually finish quickly; detailed drawings and photos can take several minutes. Each conversion has a 300-second limit; if it is not going to finish in time, the converter tells you and suggests a smaller image.

Do I need to sign in to get the files?

Converting and previewing are open to everyone; downloading the SVG, PDF, EPS or DXF needs a free Google sign-in. It is free and watermark-free either way. Anyone can run 10 conversions an hour; signing in raises that to 60 an hour and 300 a day.

How long do you keep what I upload?

Download links expire after 1 hour. The uploaded image and the generated files are removed by an automatic 1-day deletion rule and are never used for anything else. The server's working copy is thrown away as soon as the request ends, whether it succeeded or failed.

Related pages