Convert HEIF to SVG

Every other pair on this site takes a picture apart and puts the same picture back together in a different format. This one does not. A HEIF holds a grid of coloured pixels and an SVG holds shapes, and there are no shapes in the source — they have to be guessed at by finding boundaries between areas of similar colour and drawing an outline around each region.

That guess is what tracing is, and it is why the output is not the input in any sense. It also explains the whole shape of this page: whether the conversion succeeds depends almost entirely on whether your picture contains the flat regions the algorithm is looking for.

Four real files, tested first rather than afterwards. A flat logo round-tripped faithfully: 6 KB of source became 16 KB of SVG, with sharp edges, correct colours and rounded corners intact — and it now scales to a billboard, which is the case the feature exists for. A gradient banded, because a gradient is precisely the absence of flat regions.

The other two are the warning. A screenshot containing text produced an SVG forty times the size of the source, because every antialiased letter becomes its own little pile of paths. A photograph came out four times larger and posterised. Those are the numbers to weigh your own file against.

The good case is a photograph of something flat: a shop sign, a logo on a letterhead, a stencil, a stamp, an engraved plate, a drawing on paper. Those subjects genuinely are made of areas of one colour, so tracing finds real boundaries and the result is close to what a designer would have drawn.

The bad case is a scene. People, landscapes, product shots under studio lighting, anything with shadow gradation — those have no flat regions at all, so the tracer invents them by quantising the picture first. What comes back is a posterised copy in a format that is worse at holding photographs than any of the alternatives.

Because the algorithm looks for boundaries, everything that blurs a boundary hurts. Shoot straight on rather than at an angle, so a circular mark stays circular. Get even light across the whole subject: a shadow falling across half a sign becomes a colour boundary and the tracer will faithfully draw around it.

Avoid reflections and texture. A logo photographed on glossy card picks up highlights that become their own shapes, and paper grain under raking light turns into thousands of specks. Diffuse light, flat subject, camera parallel to it — the same discipline as photographing a document, for the same reason.

You choose how many flat colours the drawing is reduced to: four, eight, sixteen or thirty-two, defaulting to eight. Four is right for a single-colour mark, a stencil or a signature, and it produces the cleanest paths of any setting. Eight suits a flat illustration with a few tones.

Sixteen and thirty-two are for artwork with genuine shading, and they cost quickly — each additional colour is another set of regions with another set of outlines around them, so the file grows and the shapes multiply. On a photographed subject a high colour count mostly turns noise into paths.

The detail setting controls how tightly the paths follow the pixel edges and how small a shape has to be before it is discarded. Smooth accepts the loosest fit and throws away anything under sixteen pixels; balanced sits in the middle at eight; sharp follows closely and keeps shapes down to two pixels.

Smooth is usually the right answer for a photographed subject, because the edges in your picture are soft to begin with and the small shapes being discarded are camera noise rather than artwork. Sharp is for a clean source where every detail is deliberate — it will reproduce a ragged edge exactly as ragged as it was.

A common trick to reduce speck count is to blur the image slightly before tracing. It works, and it is not done here, because it works by softening exactly the sharp edges that are the only reason to trace a logo in the first place.

If your source needs that treatment, the tracer is telling you something about the source. A cleaner photograph, a flatter light or a lower colour count all address the cause. Blurring addresses the symptom and takes the crispness of the result with it.

Pictures larger than 1,600 pixels on the longest edge are scaled down before tracing. Tracing cost grows with the number of edges rather than the file size, and a twelve-megapixel photograph has an edge almost everywhere — left alone it can produce a multi-megabyte SVG over tens of seconds, which is a worse outcome than the one you came here to avoid.

Unlike a raster conversion, this scaling is not a loss of fidelity. The output is resolution-independent, so a logo traced at 1,600 pixels and the same logo traced at 4,000 render identically at any size. All that changes is how closely the paths hug the original pixel grid, and at 1,600 pixels that is closer than the photograph deserves.

Two corrections are applied to what the tracer emits. It writes width and height in pixels and no viewBox, which produces a file that is technically vector and practically fixed-size — dropped into a page it refuses to scale, which is the one thing you converted it for. A viewBox goes in.

It also omits the XML namespace on the root element. Inside an HTML document a browser infers it and the file looks fine, which is how the problem survives untested. Opened on its own or handed to Inkscape or Illustrator, an SVG without that attribute is not an SVG at all, so it is written in.

The page shows you the size of what you dropped and the size of what came out, and that comparison is a better verdict than any preview. A traced result several times larger than its source means the algorithm found no flat regions and manufactured thousands of tiny paths instead — no setting fixes that, because the picture is the problem.

When it goes the other way, you have what you came for: a handful of clean paths, editable in any vector tool, ready for a cutter, an engraver or a print at any size. Keep the HEIF regardless, since this site can read that format and cannot write it, and every retrace starts from the original.

Filled shapes, and only filled shapes. Each region of similar colour becomes a closed path with a solid fill, so what you open in a vector editor is a stack of coloured silhouettes rather than the layered, named, stroked artwork a designer would have drawn.

Text is the case that surprises people. Letters in the photograph become shapes shaped like letters — they are not text, they cannot be edited, respelled or restyled, and they carry no font. If the mark you traced contains a wordmark and you need to change the wording, the trace has given you a picture of the words, and the type has to be set again in the editor.

A cutter, engraver or embroidery machine follows outlines, so the useful preparation is reducing the number of them. Trace at four colours, open the result, and delete the stray shapes that came from noise or shadow — there are usually a handful, and they are the ones a cutting machine will faithfully cut.

Two more habits pay for themselves. Check that the shapes are closed and that holes in letters like A and O are genuinely holes rather than separate filled discs, because a cutter treats those differently. And set the physical size in the editor rather than assuming the trace has one: the SVG carries a viewBox and proportions, and the millimetres are your decision.