SVG upscaler: how to enlarge a PNG logo without blur
AI upscalers make a bigger bitmap that blurs again at the next size. Converting to SVG makes the size question go away. When each approach is right, and how to get a clean vector from a small PNG.
You have a 400-pixel logo and you need it on a banner, a shop sign or a 4K slide. Enlarged in any image editor it goes soft. The two fixes people search for are an AI upscaler and an SVG converter, and they are not the same thing: one makes a bigger grid of pixels, the other replaces the grid with shapes that have no size at all.
What an upscaler does
A raster upscaler, AI or classic, invents the pixels in between. Modern ones do it well: edges get sharper, textures look plausible, a 400 px image becomes 1600 px that look fine on screen. But the output is still a PNG. Scale that to a metre-wide sign and it blurs again. Print it and the printer sees the same soft edges at 300 dpi. For photos, that is the right tool. For a logo, an icon or a flat illustration, it is the wrong one.
What 'upscaling to SVG' means
A vector file does not store pixels. It stores the outline of each shape as curves and the fill inside it as a colour or a gradient. There is no resolution to run out of: the same file renders a 16 px favicon and a truck wrap. So the question 'how big can I make it' simply stops applying. The trade is that a vector only works for things that are made of shapes. A logo, yes. A photograph, no.
Why most converters make it worse, not better
The usual converter posterises the image into a few colours and traces each blob. On a small, blurry PNG it traces the blur: edges become staircases, anti-aliasing becomes slivers of odd colour, a gradient becomes stripes. The output scales to any size, but what scales is the mess.
svgize does the opposite of tracing. It measures how the source was degraded (blur, JPEG blocks, noise), then asks what design made of flat fills, gradients and clean curves would have produced those pixels. Edges are located to a fraction of a pixel and the curves are faired, so a small logo comes out as a few smooth shapes, not a copy of its blur.
How small is too small
- 200-400 px: workable for a simple mark with bold shapes. Expect to check the corners and maybe delete a speck or two.
- 500-1000 px: the comfortable range. Text in the logo comes out as clean letter shapes.
- 1500 px and up: the engine has everything it needs; fine details and thin strokes survive.
- Under 150 px, or a logo that is already a heavily compressed JPEG: look for a larger original first. An email signature, the original design brief or a social profile often has a bigger copy.
When to combine the two
If all you have is a tiny, noisy file, running it through an upscaler first can help the vectoriser: the upscaler sharpens the edges and the engine then has more pixels to measure. Convert the upscaled PNG, not the original, and compare both results at ×4. For anything 500 px or larger, skip the upscaler; it adds invented detail the vectoriser has to explain away.
After the conversion
Open the SVG in Inkscape, Figma or Illustrator and zoom to 400 %. Corners should be points, straight strokes straight, and the path count shown next to the result should be small. Export PDF for the printer, DXF for a cutting machine, or hand the SVG to the sign maker as it is.