MOS to SVG Converter

MOS to SVG: Read This First

MOS is the raw file from Leaf medium-format digital backs — sensor data in a TIFF-derived container — and SVG is XML geometry. Getting from one to the other means tracing: an algorithm groups pixels into flat colour regions and draws outlines around them. Two things follow from that, and both are worth knowing before you upload. First, a tracer needs an actual image, and a raw file is not one yet — the mosaic has to be demosaiced, white-balanced and tone-mapped before there are colour boundaries to find, so the picture the tracer sees is always a development of your raw. Developing it yourself into a clean PNG first is what puts that development under your control, and it is the largest single influence on how the trace turns out. Second, a photograph is usually the wrong thing to vectorize at all. This page covers both.

How to Convert MOS to SVG

1. Develop the MOS into a PNG First: Run your .mos through MOS to PNG. Set the Quality Preset to Highest so the intermediate is a true-colour PNG rather than a palette-quantized one — the tracer's colour clustering works better on unquantized input.
2. Upload the PNG to the SVG Tracer: Take the PNG you just downloaded and drop it into PNG to SVG. PNG is the format tracers are happiest with: lossless, so no compression artefacts get mistaken for edges.
3. Set the Number Precision: Adjust the Number precision slider. Values of 4 to 6 suit logos and flat graphics; the default is 6. Keep the slider at 8 or below — see the walk-through below for why the top of the range is a trap.
4. Convert and Download: Click "Convert" and save the SVG. Files upload over an encrypted connection, are processed on our servers, and are deleted automatically after a few hours. No sign-up, no watermark.

Walk-through: Why the Raster You Feed It Decides Everything

The tracer behind every SVG conversion here is vtracer, and it works on decoded pixels. It has no idea what a Leaf back is; what it does is scan an RGB image, cluster adjacent pixels whose colours are close enough, and fit Bézier curves around each cluster. Everything about the output — how many paths, how clean the outlines, how large the file — is a consequence of the pixels handed to it.

A raw file is a poor thing to hand it directly, because a raw file is not RGB pixels. Its data is a colour-filter-array mosaic: one filtered brightness value per photosite, which has to be demosaiced into red, green and blue, then white-balanced, then run through a tone curve before any of it is an image. Somebody makes those choices. If you develop the file yourself with MOS to PNG, you make them — you can check the exposure, avoid crushing shadow detail into one undifferentiated black region, and downscale a 60-megapixel frame to something a tracer can find real edges in rather than sensor noise. Development on defaults does none of that for you, and every one of those decisions shows up in the path count.

There is a second trap worth knowing about, and it is independent of anything else on this page. The Number precision slider on the SVG pages runs from 1 to 10, and its value is passed to the tracer as both the path precision and the colour precision. Colour precision is only valid from 1 to 8. Set the slider to 9 or 10 and the tracer aborts with Out of Range Error: Color precision is invalid at 9. It must be within [1,8]. Keep it at 8 or lower and you will never see it.

What Makes a Good Tracer Input

Choosing a Number Precision

Common Errors and How to Fix Them

• "No image file found at specified input path" — The tracer was handed something it could not open as an image. The usual cause is a file that was renamed rather than converted: giving a raw a .png name changes the label, not the sensor mosaic inside, so a PNG decoder finds no PNG there. Run a genuine render instead — MOS to PNG — and trace the result.
• "Color precision is invalid ... It must be within [1,8]" — The Number precision slider is at 9 or 10. Drop it to 8 or below and re-run.
• "The SVG looks flat and posterized" — Expected. Tracing replaces continuous tone with regions of solid colour. If you wanted the photograph to still look like a photograph, you want a raster target: MOS to PNG or MOS to JPG.
• "The SVG is far larger than the source image" — A detailed photograph forces the tracer to emit an enormous number of paths, most of them describing background texture. Lower the Number precision, or downscale the intermediate PNG before tracing, or accept that a raster format is the right answer for this image.
• "My editor opens the SVG but the paths are unusable" — Photographic traces produce thousands of overlapping shapes that no one can edit by hand. Trace a simplified or higher-contrast version of the image, or reduce the palette on the intermediate PNG first so the tracer has fewer regions to find.

When This Doesn't Work

Even with a perfectly prepared PNG, SVG is the wrong destination for most MOS files. A Leaf back exists to record smooth gradients and fine tonal transitions across 22 to 80 megapixels — precisely the content vector tracing cannot represent. What comes back is a stylized posterization with a huge path count, not a scalable version of your photograph. The conversion is worth doing when the frame is logo-like: a photographed sign, a signature, a scanned drawing, a high-contrast graphic, artwork you intend to re-colour as flat shapes. For anything else, convert to PNG, JPG or TIFF and stop there. There is also no way to "trace back" to a raw: vectorizing is a one-way, heavily destructive transformation, so keep the .mos regardless of what you do next.

Frequently Asked Questions

Why develop the MOS to PNG instead of just tracing the raw?

Because a raw file's pixel data is a single-channel mosaic from the sensor's colour filter array, not an RGB image. It has to be demosaiced and white-balanced before there are colour regions to cluster at all, and whoever performs that step decides the exposure, the colour and the noise floor the tracer then sees. Doing it deliberately on MOS to PNG means you can inspect the render, fix a bad exposure, and downscale before tracing — three things that change the output far more than any tracing setting does.

Will converting MOS to PNG first lose anything that matters for tracing?

Practically, no. Tracing already discards far more than the PNG step does — it collapses millions of colours into a handful of flat regions. Use the Highest preset on MOS to PNG so the intermediate is true-colour rather than palette-quantized, and the tracer will see essentially everything it is capable of using.

Can I use JPG or TIFF as the intermediate instead of PNG?

Yes — MOS to JPG followed by JPG to SVG is a perfectly workable route. PNG is still the better intermediate, because JPEG compression adds ringing around hard edges, and hard edges are exactly what the tracer is looking for — JPEG artefacts become spurious little paths in the output, inflating the path count for detail that was never in the photograph.

What does Number precision actually control?

It sets how many decimal places the tracer keeps in each path coordinate, and it simultaneously sets how finely colours are clustered. Lower values give fewer, blockier shapes and a smaller file; higher values keep more detail and inflate the path count. Because the same slider drives colour precision, which the tracer caps at 8, values of 9 and 10 fail outright rather than giving you extra detail.

Is there any setting that makes a photograph trace well?

No, and no tracer has one. The limitation is structural: SVG describes shapes with defined edges, while a photograph is a field of continuously varying values with no edges in the geometric sense. You can make a photographic trace look deliberate by reducing the palette first so the posterization reads as a design choice, but you cannot make it look like the photograph.

Does the SVG output scale to any size without quality loss?

Yes — that part of the promise is real. Once the trace exists, the paths are resolution-independent and render crisply at any size, which is why the format is worth the trouble for logos and line art. What scales infinitely is the traced approximation, though, not your original image; the fidelity was decided at trace time and no amount of scaling recovers it.

Which apps can open the SVG I get back?

Any of them. The output is standard SVG, so Inkscape, Adobe Illustrator, Affinity Designer, Figma, and every current browser open it directly. If you plan to edit the paths by hand, trace at a lower Number precision — fewer, simpler paths are far easier to work with than a technically more accurate mesh of thousands.

What happens to my files after conversion?

Every file is uploaded over an encrypted connection, processed on our servers, and deleted automatically after a few hours, at both stages of the route. There is no sign-up and no watermark, and nothing you upload is shared or made public.