Rasterless

Turn a JPG into a real vector file

Two jobs in one extension: posterising a photo, or rescuing a compressed logo.

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 shaded badge JPEG on the left, the same badge traced into flat vector colour regions on the right — source image
A shaded badge JPEG on the left, the same badge traced into flat vector colour regions on the right — vector result

01 · 5 colours · 193 nodes

How it works

  1. Upload your JPG — a photo, a scan, or a screenshot of a logo. PNG and WEBP work too; one image per conversion, up to 25 MB and 40 megapixels, and anything above 1.5 megapixels is downscaled before tracing.
  2. Decide which job this is. For a photograph start at 12 colours; for a logo, set the number of ink colours you can count in the mark, usually 2 to 6.
  3. Choose outputs: SVG to keep editing, PDF or EPS for a print shop, DXF for cutting machines, PNG for a cleaned-up raster. Pick several and you also get a ZIP.
  4. Add a width and height in millimetres if the piece has a known physical size, then check the 800 px preview.
  5. Sign in with Google (free) to download the files.

A JPG is rarely clean artwork. It is what a camera writes, what a phone shares, and what you get when the only copy of a logo is a screenshot of a web page. So "JPG to vector" is two different jobs wearing one file extension: flattening a photograph into vector shapes, and rescuing flat artwork that lossy compression has chewed up.

They want opposite colour counts, so this page keeps them apart.

Take the photograph first; that is the expectation worth correcting. Tracing a photo does not hand you a photograph made of curves. The tracer segments the picture into a fixed number of flat colour regions: shading turns into bands, skin becomes three or four tones, a soft background collapses into blobs. What comes out is posterised and stylised — the look you want on a shirt, a poster or a cut stencil, enlarging to poster scale without pixel edges, but not a faithful reproduction. If the photograph has to stay photographic, keep the raster.

The logo case runs the other way. That artwork was already flat; JPEG is what damaged it. Lossy encoding works on 8×8 pixel blocks and discards colour detail faster than brightness, so a hard edge returns wrapped in a faint ringing halo and a solid red field comes back mottled. Left alone, the tracer faithfully vectorises that damage. Told how many colours the mark actually has, it steps straight over it.

What actually happens

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

Where JPEG compression ends up in the vector

Colour edges blur more than brightness edges in a JPG

Most JPGs store colour at half the resolution of brightness, so an edge between two tones of similar lightness is smeared over three or four pixels while a dark-on-light edge stays crisp. The tracer places both where the ramp crosses halfway, but the wider ramp gives it more room to wander. Expect dark outlines to come back close and same-brightness colour boundaries to need a glance.

Re-saving a JPG at higher quality restores nothing

Compression damage is cumulative and one-directional. Opening a JPG and saving it again, even at the highest quality setting, re-encodes the blocks and adds a second layer of ringing on top of the first; nothing comes back. The same goes for a screenshot of the logo on a web page, which adds the browser's scaling as well. Trace the earliest copy you can find, untouched.

A busy JPG background steals colours from the face

The colour count is shared across the whole frame. Segmentation picks the twelve tones that cover every pixel most closely, and a busy background, foliage or a patterned wall, claims most of them, leaving a face with two. Crop the photograph to the subject before converting and the same twelve colours go where you wanted them; a crop of a large photo may also escape the 1.5 megapixel downscale.

Two colours ignore the JPEG halo; more colours trace it

At 2 colours the compression halo falls to one side of the edge or the other, so a stencil cut from a JPG comes out about as clean as one from a PNG. The trouble starts the moment the count exceeds the colours the mark really has: the halo is then the next most distinct thing, earns a cluster of its own, and traces as a thin ring around every letter.

Choosing the colour count

This page is for people who did not get to choose the source file. A designer inherits a client's logo as a JPEG pulled off an old website. A print shop is handed a snapshot and asked for spot colours. A craft cutter wants a silhouette out of a holiday photo. A print-on-demand seller needs artwork that survives enlargement.

Vectors solve the scaling half: curves and closed shapes stay crisp at whatever size you output, and colours become objects you edit rather than pixels you retouch. They cannot invent detail the JPEG no longer holds.

One image goes in per conversion; accepted inputs are PNG, JPEG, WEBP, BMP, GIF and AVIF. The engine combines geometric tracing (colour segmentation, an explicit anti-aliasing model and curve fitting) with machine learning (AI). Even so, the colour count matters more than anything else. It is free and watermark-free; you can run a conversion and study the preview first, and downloading asks for a free Google sign-in.

Settings we suggest

Colours decides everything, and the right number differs completely between the two cases.

For a photograph, start at 12. Six to ten gives a bold screen-print look with clean separations; twelve to sixteen keeps a face or a product recognisable; past twenty-four you are spending file size on slivers of compression noise, and it still will not read as a photograph. Two colours turns the shot into a black-and-white stencil — the right answer for vinyl and laser work.

For a logo recovered from a JPEG, count the ink colours in the mark and set that number, plus one if a background colour is worth keeping. Most marks land between two and six; a single-colour wordmark belongs at 2. Reaching for a high number is the classic mistake: it hands you the compression halo as extra shapes around every letter. The range is 2 to 64.

Leave draw style on Fill shapes for both cases; Stroke edges is for contours you will follow with a pen or a blade. Shape stacking on Cut-outs gives every colour its own closed region, with holes where another sits on top — what screen-print separations and vinyl need; Stacked paints them over each other instead. Group by colour puts each tone in its own group, so a separation selects at once.

DXF is R2010 (AC1024) and holds outlines only, LWPOLYLINE and SPLINE, one layer per colour; it opens in AutoCAD 2010 and later, Fusion 360, LightBurn and Inkscape, and it states millimetres in its header, so the machine opens it at size rather than guessing the unit. The millimetre field is written into all four vector formats — geometry in the SVG and DXF, page box in the PDF and EPS; leave it empty and 96 px = 25.4 mm, with the PDF and EPS page sized from the pixel dimensions.

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.

Questions

Will my photo still look like a photo after I vectorise it?

No. Continuous shading becomes a fixed number of flat regions, so you get a posterised, stylised version: excellent on shirts, posters, stickers and stencils, but not a faithful reproduction. Keep the raster if the photograph itself is the deliverable.

How many colours should I use for a logo I only have as a JPEG?

Count the ink colours in the mark and use that number, plus one for a background worth keeping. Two to six covers most logos; a single-colour wordmark belongs at 2. A high count cannot recover lost detail — it vectorises the compression halo instead of the letters.

Why does my traced JPG have speckles and coloured fringes at the edges?

That is the compression, not the tracer. Lossy encoding works on 8×8 pixel blocks and keeps less colour detail than brightness, leaving ringing beside hard edges and mottling inside flat fields. Lowering the colour count merges those artefacts into the surrounding colour.

My JPG has a white background — can I get a transparent vector?

JPEG carries no transparency channel, so that white is real pixels, traced like any other colour. At higher counts it arrives as a white shape you can select and delete in a vector editor; at 2 colours you usually get a single silhouette contour.

What happens to the JPG I upload?

Download links expire after 1 hour. Uploaded images and generated files are removed by an automatic 1-day deletion rule and are never used for anything else. Anonymous use is capped at 10 conversions per hour; signed-in users get 60 per hour and 300 per day.

Related pages