Rasterless

Turn the image you already have into an SVG

Whatever the extension, the picture's kind — logo, sticker, photo, line art — sets the trace.

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 four-colour badge — dark teal frame, teal field, cream triangle and orange triangle — shown as the original raster picture and as the traced SVG — source image
A four-colour badge — dark teal frame, teal field, cream triangle and orange triangle — shown as the original raster picture and as the traced SVG — vector result

01 · 5 colours · 193 nodes

How it works

  1. Drop in the picture — PNG, JPEG, WEBP, BMP, GIF and AVIF are the accepted types, and the bytes decide, not the extension.
  2. Name the kind of picture you have: flat logo, sticker or badge, photograph, or line drawing.
  3. Set the colour count from that kind — 2 for a silhouette, 4–10 for a flat logo, 12–20 for a sticker, 24–48 for a photograph.
  4. Leave Draw style on Fill shapes and Shape stacking on Cut-outs unless you want contours only.
  5. Open the preview at full size, then sign in with Google (free) to download the SVG.

Most people who need an SVG do not arrive thinking in file formats. There is a picture — a badge a designer emailed over, a sticker saved off a phone, a mark screenshotted from an old slide deck — and it has to become a vector before it can be cut, printed or scaled. The extension is the least interesting part of it.

The raster types that come through here: PNG, JPEG, WEBP, BMP, GIF and AVIF.

Working out which one you are holding takes a moment: Properties on Windows, Get Info on a Mac, the details sheet on a phone. When the name and the contents disagree — a photo a chat app handed you as logo.png that is really JPEG data inside — the file's own signature settles it and the conversion runs anyway. A GIF goes in as it is — an animated one is traced from its first frame only. An iPhone HEIC or an old TIFF scan is a different story: re-save it as PNG in any editor first.

What genuinely changes your result sits one level above the format. A two-colour silhouette, a flat logo, a sticker with soft shading, a photograph and a pencil line drawing each want a different number of colours, and that number has nothing to do with whether the bytes arrived as PNG, JPEG, WEBP, BMP, GIF or AVIF. Sort your picture into one of those kinds and the settings follow.

What actually happens

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

Why the picture's kind outranks its file extension

Decoding erases the extension; one pixel grid remains

PNG, JPEG, WEBP, BMP, GIF and AVIF are decoded to the same thing: an RGB grid plus, for PNG and WEBP, an alpha plane. From there every stage — Lab clustering, ramp dissolving, the sub-pixel edge, the curve fit — is identical, and nothing branches on the container. What the container leaves behind is damage, or its absence: JPEG and lossy WEBP leave halos, PNG and lossless WEBP leave none, and that is what the colour count has to absorb.

Faint boundaries stay on the pixel grid

The sub-pixel step slides each boundary to the halfway point of the anti-aliasing ramp between its two colours, and it needs a measurable ramp to do so. Where two regions differ by less than about three levels in RGB there is nothing to measure, and the boundary stays on the pixel crack it was found on. Pastel stickers and low-contrast photographs are where this shows: edges between near-identical tones keep a faint stair-step that stronger edges lose.

Line art wants the darker side drawn, once

Choose Stroke edges and each boundary is written a single time as an open path in the darker of its two colours — an edge between white paper and black ink is drawn black, never the pale mean that would vanish on the page. Strokes are 2 px wide with round caps and joins, and the fills are dropped. That is the right output for a pen plotter or an engraver, and the wrong one for anything that needs an area.

Group by Parent or Layer follows the picture's nesting

Colour gathers everything of one colour into a single selectable object, which suits a logo you mean to recolour. Parent writes one g per containing shape and Layer one g per nesting depth, so a sticker becomes layer0 for the field, layer1 for the shapes inside it, and so on — a frame and its contents selectable apart. Grouping changes the markup only; the geometry inside is identical whichever you pick.

Choosing the colour count

This page is for when the destination is fixed but the source is whatever you happen to have. A vinyl cutter wants clean closed paths; a print shop wants artwork that stays sharp at A2; a website wants a mark of a few kilobytes that holds up on a retina screen. In each case the answer is an SVG built from real path data, not the original bitmap wrapped inside an SVG tag.

Vectorising is a re-drawing, not a recovery. A 200 px wide, soft original gives a faithful vector copy of something soft: crisp edges, same mushy shapes. Start from the largest copy you can find, and prefer one that never went through a messaging app.

The engine combines geometric tracing — colour segmentation, an explicit anti-aliasing model, curve fitting — with machine learning (AI) that runs on our own servers. Every control is still one you can reason about: how many colours, fills or strokes, how the shapes stack.

Settings we suggest

Name the picture first, then set the colour count. If you cannot tell, leave it on Auto (it uses 26 when it cannot read a count) and judge the preview at full size.

Flat logo or wordmark. Count the flats the design already has, including a solid background, and use that number — usually 4–10. Too many colours invents slivers along edges where the original had anti-aliasing; too few merges two brand colours into one.

Sticker or badge. These look flat but usually carry two tones of one hue, a shadow and an outline; 12–20 keeps the shading readable. The sample here is that case: its teal frame and teal field are two colours the eye reads as one.

Photograph or painted artwork. No correct number exists, only a poster effect you choose: 24–40 for a printable poster, 48–64 to keep more of the tonal roll-off at the cost of a heavier, slower file.

Line drawing or silhouette. 2 colours produces one clean shape against nothing at all, which is what a cutter or an engraver expects; grey strokes rather than black ones need 3–4 so the lighter marks survive.

The rest of the panel barely moves between kinds. Keep Draw style on Fill shapes unless you want contours only, and Shape stacking on Cut-outs so every colour region is a closed shape with its holes cut out. Group by colour collects each colour into one SVG group.

Transparency is honoured: fully transparent pixels stay empty in the vector, semi-transparent ones are composited onto white. Set a size in mm if the SVG must land at an exact physical size; otherwise it carries pixel dimensions, where 96 pixels = 25.4 mm.

Questions

How do I tell whether my file is a PNG, a JPEG or a WEBP?

Properties on Windows, Command-I on a Mac, the details sheet on a phone. Or just try it — the converter reads the bytes at the start of the file rather than the extension, so a JPEG that somebody renamed to .png still goes through.

My picture is a HEIC from an iPhone. Can I still get an SVG?

Yes, after one extra step: re-save it as PNG. Exporting the photo, or emailing it to yourself, usually produces a PNG or JPEG on its own; in a desktop editor, use Export As. The same goes for an old TIFF scan. A BMP or a GIF goes in as it is; an animated GIF is traced from its first frame.

Does it matter which of the accepted formats I start from?

Less than people expect, since the tracer works from decoded pixels. A PNG, or a losslessly saved WEBP, is cleanest. A JPEG — and a lossy WEBP — spreads faint halos around hard edges that can turn into thin extra shapes, so step the colour count down a notch. PNG and WEBP also carry transparency; JPEG does not.

What are the size limits, and how long does a conversion take?

Up to 25 MB and 40 megapixels, one image per conversion; anything above 1.5 megapixels is downscaled before tracing. 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.

Is it free, and what happens to my picture afterwards?

Free, and nothing is watermarked. The preview is open to everyone; a free Google sign-in unlocks the vector download. You get 10 conversions per hour before signing in, 60 per hour and 300 per day after. Download links expire after 1 hour, and your files are removed by an automatic 1-day deletion rule.

Related pages