Photo to STL Converter

This free photo to STL converter turns one photo (up to 10 MB) into a printable relief, lithophane or silhouette STL in your browser, with nothing uploaded. A single photo cannot become a full 3D model: the result has a flat back, because the camera never recorded the sides or rear — full models need photogrammetry, CAD or the separate paid AI mode, which only guesses the hidden side.

Last updated: 2026-10-22

PNG · JPG · WEBP → STL

Two very different things are called photo to STL

The search phrase covers two products that behave nothing alike. One is a relief: brightness becomes height, the back stays flat, and you print a shallow 3D version of the photo that is meant to be seen from the front. The other is reconstruction: software guesses a complete object with sides and a back from one or more images. Confusing the two is why people are disappointed with tools that were never built for the job.

Relief is honest and deterministic. A heightmap is a direct transform of pixel brightness into Z height on a solid base, so what you see in the preview is what you print. It works with any photo, needs no account, costs nothing and produces good results on the first attempt once contrast and crop are right.

Reconstruction is inference. Photogrammetry needs many views of a static object; single-image AI invents the unseen side from learned priors. This site offers that as a separate paid mode — it uploads the photo, needs an account and uses credits per generation. It is good for stylised figurines and fun, and it is not a measuring tool.

A blunt warning worth stating: no single photo can recover what the camera never recorded. If the back of the head or the far side of a part was never in frame, an AI result is a plausible guess, not an observation. For brackets, replacement parts or anything that has to fit, use CAD or photogrammetry with full coverage, not a single photo.

Free route: photo to relief or lithophane

Drop the photo, then choose what you want to end up with. A relief uses heightmap mode with bright areas high, which reproduces the lighting of the original as physical depth. A lithophane uses the same mode with invert enabled, so dark areas become thick: the image only appears when light shines through the print.

For a portrait relief, start at 100 mm on the longest side with 3 mm of relief, then push the strength to 5 or 8 mm if the face reads flat in the preview. Keep the base at 1 mm for a plaque that can stand or be glued; drop it to 0.6 mm if you want a lithophane thin enough to transmit light.

For a silhouette cut-out — a head, a pet, a shape you want as a flat badge — switch to extrude mode and let the threshold decide what stays solid. That is the same mode as a logo conversion: what matters is a clean, high-contrast boundary between subject and background.

Choosing and shooting the photo

Contrast beats resolution every time. A 1000-pixel photo with strong directional light and real shadow converts into a relief with visible facial structure; a 24-megapixel photo taken in flat overcast light converts into a nearly flat panel. If you can reshoot, do it — one lamp at 45 degrees changes the result more than any slider.

The background decides the composition. Because the photo is stretched across the model, a wide background shrinks the subject to a bump in the middle, while a tight crop across the shoulders uses the whole panel. Crop close, leave a small margin, and make sure the background is either uniform or removed.

Resolution beyond roughly 1500 pixels on the long side buys nothing, because images are resampled to a 256-cell grid before geometry is built. That also means a phone photo is plenty — and the 10 MB file cap is easy to stay under if you do not upload a raw export.

Photo to STL problems and fixes

A face that looks like a crater is the classic inversion mistake: bright areas sit low, shadows sit high. Toggle invert to flip the mapping, or change the lighting assumption of the source photo — reliefs read correctly when highlight means high.

A muddy panel with no structure is low contrast or a lit-from-the-front photo. Increase relief depth, raise contrast in the source, or pick a shot with a shadow across the features. Smoothing helps noise but cannot invent depth that was never in the luminance range.

Hair, lace and other fine textures are the first thing to disappear, because each cell of the grid is about 0.4 mm at a 100 mm width and anything thinner than that cannot survive. Either accept a softer result, print larger, or move up to the AI mode where the surface comes from a mesh rather than a depth map.

Print settings matter as much as conversion settings. A relief printed flat on the plate gets a mirror-smooth back and a top surface with layer lines running across the face; printed upright it needs a brim and rear supports, and the layered surface then faces sideways where raking light hides it. Layer height around 0.12 mm keeps facial detail readable, while 0.2 mm looks noticeably coarser. Resin beats FDM for small portrait reliefs; a 150 mm panel on FDM at 0.12 mm with a 0.4 mm nozzle is usually good enough.

Photo to STL FAQ

Is photo to STL free here?

The relief and lithophane tools are free, unlimited and require no account, because they run in your browser on your own hardware. AI reconstruction from a single photo is a separate paid mode that uploads the image and consumes credits — nothing on this page forces you into it.

Is my photo uploaded?

In the free modes, no. The file is decoded locally, sampled into a grid and turned into geometry in your browser, and the STL you download was generated on your machine. Only the optional AI mode uploads the image, and it says so before you start.

What is the difference between a relief and a 3D model?

A relief has real height but a flat back and no side or rear detail — it is a 3D surface, not a 3D object. A reconstructed model has geometry all round. Relief is generated deterministically from brightness; a full model is inferred, either from many photos or, in the AI mode, guessed from one.

Which photos convert best?

Single subject, tight crop, plain background, strong directional light. Portraits with a window or lamp to one side, pets photographed in daylight, and product shots with a clear shadow all convert well. Flat flash lighting, busy backgrounds and low-contrast scenes are the three biggest causes of disappointing results.

How large can I print the result?

The default longest side is 100 mm, and you can set it lower or higher before converting. Scaling up in the slicer also works, since STL is unitless, but remember that detail is limited by the 256-cell grid: a bigger print shows the same amount of detail, just larger and softer.

What settings make a good lithophane?

Heightmap mode with invert on, a base around 0.6 mm, relief deep enough for a visible thickness range, then print upright at 100 percent infill with white or translucent filament at 0.1 mm layers. Warmer, brighter light shows more contrast; don't judge the print until it is backlit.

Can two photos be combined into one model?

Not in the free tool: one image becomes one relief. Combining views is photogrammetry or AI reconstruction territory, and it needs overlapping coverage of the whole subject. If you only have a front view, build the relief and accept that the back is a flat plate.

Can one photo become a full 3D model?

Not with the free tool. One photo records one viewpoint, and a full model needs the sides and back the camera never saw. Heightmap and extrusion give a relief with a flat back — dimensionally honest, and the right output for lithophanes, plaques and portraits. The paid AI mode infers a full model from a single photo at 20 credits per generation, but the hidden side is a guess, so check it before printing anything that has to fit.