MRW to SVG Converter

MRW to SVG: Read This First

Straight answer, because it saves you a disappointing result: the trace is only ever as good as the raster it is given, and a Minolta RAW is not a raster. The vector tracer behind this page works on decoded RGB pixels — it clusters neighbouring pixels into flat regions and fits curves around them. An .mrw holds a Bayer mosaic that has to be demosaiced, white-balanced and tone-curved before there are any colour boundaries to find, and whoever does that development chooses the exposure, the colour and the noise floor the tracer then sees. Developing the RAW to PNG yourself first, and tracing that, is the route that puts those choices in your hands.

This page explains that route, and — more usefully — helps you decide whether tracing a photograph to SVG is a good idea in the first place. Usually it is not. SVG stores shapes, not pixels, so vectorising a 12-bit sensor capture of a real-world scene means approximating millions of subtly different colours with thousands of filled paths.

How to Convert MRW to SVG

1. Develop the MRW to PNG First: Run your file through MRW to PNG and set the Quality Preset to "Highest". That preset skips the palette quantiser, so the tracer sees the full render rather than a reduced palette.
2. Upload the PNG to the Tracer: Take the resulting .png to PNG to SVG. PNG is the ideal tracer input — lossless, so there are no compression artefacts for the edge finder to mistake for real detail.
3. Set Number Precision: The single slider runs from 1 to 10 and starts at 6. Keep it at 8 or below — see the table on why 9 and 10 abort the job.
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.

Why the Intermediate PNG Is Worth the Extra Step

The tracer is a raster-to-vector program, not a photo pipeline. It has no view on exposure, no idea what a Minolta sensor's colour response looks like, and no way to recover a blown highlight. It takes whatever RGB pixels it is handed, groups them into flat colour regions, and draws outlines. Everything about the result — how many paths you get, whether the outlines follow the subject or the noise, whether the file is small enough to edit or too heavy to open — is decided by those pixels.

That is why a deliberate development pass matters. Rendering the RAW yourself gives you a chance to look at the frame, correct an exposure that would otherwise crush the shadows into one shapeless black region, and downscale a full-resolution capture so the tracer finds edges rather than sensor grain. A RAW developed on default settings gets none of that attention, and the difference shows up directly in the path count. Renaming the file achieves nothing either way — the bytes inside stay a sensor mosaic whatever extension they carry.

Number Precision, and Why 9 and 10 Break

One slider on the tracer page carries two jobs: it sets colour precision (how many bits per channel the engine uses when clustering pixels into shapes) and coordinate precision (decimal places in the path data). The slider's range runs to 10, but the tracer only accepts colour precision values from 1 to 8.

If a trace stops without producing a file and you have pushed the slider to the top, that upper limit is the first thing to check. Drop it to 8 and rerun.

When Tracing a Photo Is the Wrong Tool

Vectorising works when the source is already made of flat regions — a scanned logo, a line drawing, a screenshot of a chart, a signature. A photograph from a DiMAGE or a Maxxum 7D is the opposite: continuous tone, sensor noise, and gradients everywhere. The tracer has to fabricate a path boundary for every colour step, so the SVG ends up both enormous and less faithful than the raster it came from.

If you want a scalable version of an old Minolta photograph, the better move is almost always to keep it as a raster and simply export at higher resolution. Reach for a trace only when the picture genuinely contains flat artwork you want as editable shapes.

Frequently Asked Questions

Why bother developing to PNG rather than handing the RAW straight over?

Because the RAW has to be developed either way — the only question is whether you make the decisions or a default profile does. A .mrw stores one filtered brightness value per photosite, not RGB pixels, so demosaicing, white balance and a tone curve all have to be applied before there is anything for a tracer to cluster. Doing that step deliberately on MRW to PNG lets you check the exposure, keep the shadows separated instead of clipped into one black mass, and pick a sensible pixel size. Those three things move the output far more than the tracing slider does.

Will converting via PNG lose anything worth keeping?

Practically nothing that tracing would have used. Vectorising already collapses millions of colours into a handful of flat regions, so a lossless PNG in the middle costs you nothing the tracer could have exploited. The one thing to get right is choosing the "Highest" preset on the PNG step, since the other presets reduce the image to a colour palette before the tracer ever sees it, which changes what shapes the tracer finds.

How big will the SVG be?

Larger than you expect. Tracing a 720 × 480 photographic test image at the default precision of 6 produced a roughly 1 MB SVG in our own run — several times the size of the PNG it came from, and that was a small image. A full-resolution frame from a Maxxum 7D is roughly seventeen times as many pixels. Expect very large files and slow rendering in browsers, which is the practical argument against tracing photographs.

Can the SVG be edited in Illustrator or Inkscape?

Yes — the output is ordinary SVG path data, so any vector editor will open it. Whether it is workable is another question. A traced photograph arrives as thousands of stacked filled paths with no meaningful grouping, so editing it feels less like working with vector artwork and more like sorting confetti. Traces of flat artwork are far more tractable.

What is MRW, and why is it so awkward to open?

MRW is Minolta's RAW container, written by the DiMAGE 5, 7, 7i and 7Hi, the DiMAGE A1, A2 and A200, and the Alpha/Dynax/Maxxum 5D and 7D. Konica Minolta announced its exit from the camera business in January 2006 and transferred SLR operations to Sony that March, so the format was orphaned and mainstream support has been eroding ever since. Open-source RAW decoders still read it, which is how the PNG step in this workflow works.

Does SVG keep the photo's colours accurately?

Approximately at best. The tracer clusters similar colours together to form shapes, which is a deliberately lossy simplification — smooth gradients become visible bands of flat fill. Raising the precision slider narrows the bands but multiplies the path count and file size. There is no setting at which a traced photograph becomes colour-accurate; that is inherent to representing continuous tone as filled regions.

I want a scalable version of an old photo. What should I actually do?

Export a high-resolution raster instead of vectorising. MRW to PNG at the Highest preset with resolution left on "Keep original" gives you every pixel the sensor recorded, which is the real ceiling on detail — no trace can invent more. If you need it larger than the sensor captured, an upscaler will serve you better than a tracer, because tracing does not add information, it discards it and then draws smooth edges around what is left.

How are my uploaded files handled?

Files are uploaded over an encrypted connection, processed on our servers, and deleted automatically after a few hours. No sign-up, no watermark, and files are never shared or made public. That applies to both steps of the workflow — the RAW you upload and the intermediate PNG you upload afterwards.