Vectorize an image, in your browser.
See the anchor points before you download. Free, and nothing is uploaded.
Drop a PNG, JPG or WebP
or paste from the clipboard
Best from 500 px wide, sharp edges, plain background
Or try:
See the anchor points before you download. Free, and nothing is uploaded.
Vectorizing is the process of turning a picture made of pixels into one made of shapes. Most tools ask you to take the result on faith. This one shows you the geometry: switch on Paths and every outline and anchor point is drawn over the artwork, on screen, before you download anything.
Six steps, all of them on your own computer:
The how it works page goes through each step in detail if you want the mechanics.
Most people look at the vector, decide it resembles the original, and download. Two extra checks take ten seconds and catch nearly every problem.
There is also Split view, which puts a draggable divider across a single frame so you can compare the original and the vector at exactly the same spot rather than glancing between two panels.
Every change re-traces from your original image, so nothing degrades through repeated adjustment and you can move a slider back and forth without accumulating damage.
The process needs edges to follow. Artwork built from flat colour with clear boundaries gives it plenty: logos, wordmarks, icons, signatures, hand lettering, line drawings, stencils, scanned sketches, flat illustration.
Photographs do not, and this is worth saying plainly rather than letting you discover it. A photograph is continuous tone with no flat regions, so the tracer has to invent boundaries. You get a posterised approximation of hundreds of shapes, larger than the original and awkward to edit. If a photograph genuinely has to become vector artwork, a machine-learning tracer such as Vectorizer.AI is designed for it and will do better.
Related: image to vector, image to SVG, and the guide to vectorizing an image step by step.
<path> elements with cubic Bézier curves.<g fill="#…">, so one click selects a whole colour in Illustrator, Figma or Inkscape.<linearGradient>.It looks for boundaries between areas of similar colour, follows each boundary as an outline, and fits smooth Bézier curves along it. The result describes your artwork as shapes and colours rather than as a grid of pixels.
Usually about a second, sometimes a few seconds for a large or detailed picture. It runs on your own computer rather than in a queue on a server, so the speed depends on your machine and not on how many other people are using the site.
No. Your file is read into memory and never modified. Nothing is written to your disk unless you click download, and the file is not uploaded — unless you deliberately attach it to a report of a poor result.
Because the process depends entirely on there being edges to find. Flat colour with crisp boundaries gives the tracer something unambiguous to follow. Soft gradients, blur, noise and photographic texture give it nothing definite, so it either invents boundaries or misses them.
Each one sets the whole pipeline for a kind of artwork: how much to enlarge before tracing, how aggressively to smooth, how small a shape may be before it is discarded, and whether texture is kept. GetSVG picks one by profiling your pixels, and you can override it at any time.
Start with the number the artwork genuinely has. A two-colour logo needs two, not sixteen. Every extra colour becomes another layer of shapes, which makes the file larger and the result harder to edit without looking better.
Nothing broke; the tracer is following detail that is really there, usually JPEG noise, paper grain or soft shading. Move the Detail slider towards Simpler, raise Remove specks, or reduce the colour count. Watch the shape count in the stats bar as you do.
Move the setting back and it re-traces from your original image every time, so nothing degrades through repeated adjustment. The Reset link beside Fine-tune returns everything to the preset's defaults in one click.