From photo to 3D model: try it free, then print

How do I make a 3D image from a photo?

A photo is flat; a 3D image has depth, and that depth has to come from somewhere. An AI generator estimates what the object looks like from every side, a relief tool translates light and dark into high and low, and a lithophane tool translates light and dark into thin and thick. The three routes side by side:

AI: a spatial model, nearly right

The AI designer looks at one to 4 photos of the same object, or at a description in words, and estimates what the object looks like from every side. The result is a real 3D model that you rotate in the viewer and then download as STL, 3MF or OBJ for printing, or as GLB with colour. It usually takes a few minutes, longer when busy.

You pay with credits: standard 1 credit, premium 4 credits per design. If a design fails, the credit comes back; it costs you nothing. The model is yours. To try it, you receive welcome credits after verifying your e-mail, while that month's supply lasts.

The line is: nearly right, not a copy. The shape and the proportions are usually correct, but fine details, text and the back are an estimate. A face never becomes a true likeness, and the model is not dimensionally accurate. For a figure or a design you still want to work on yourself, that is enough. For a part that has to fit, or a portrait that really has to look like someone, the modelling service is the route: a modeller then works from dimensions and photos, with a quote first.

If you want more control, you take the same model into the studio: there you pick a base, measure in millimetres, make variants and keep versions, with the same credits.

Relief: the photo as a height map

In a relief every pixel becomes a height: light becomes high and dark becomes low, or the other way round. You get a plate on which the image can be felt. That works well for logos, lettering, coins and profile portraits, and badly for photos with a lot of depth, because a relief does not know what is in front and what is behind. A face seen from the front becomes a mask, not a head.

You can make a relief for free in the browser, for example with ImageToStl.com: you upload an image and choose a height; black becomes zero, white becomes the full height. Images larger than 1000 by 1000 pixels are reduced there. The file comes back as an STL, which you can simply have printed.

Lithophane: only visible with light behind it

A lithophane is a thin plate whose thickness varies from point to point. In ordinary light you see a vague grey plate; hold it in front of a lamp and the photo appears, because thin parts let more light through. It is a classic gift, but not a spatial object. It only works with a light source behind it and in one colour, usually white.

There are free generators for this too. ItsLitho calls itself free online design software for lithophanes, with flat, curved and round shapes and a preview with light behind it. 3dp.rocks is a free, open source lithophane maker in the browser with a switch between negative (lithophane) and positive (relief). Both deliver an STL.

How can I make a 3D model from photos for free?

There are three free routes, and they ask something different of your photos. Photogrammetry wants many photos all around the object and calculates the shape from them; AI wants one to 4 photos and estimates the rest; relief and lithophane want one high-contrast photo and turn it into a plate.

Photogrammetry really measures: what you photograph comes back in the right proportions, including the back, as long as you photographed it. In return you have to be able to walk around the object, shiny and transparent things work badly, and the result often still needs cleaning up. AI asks for one photo and gives a closed model within minutes, but it estimates. For a quick 3D model from a photo of your favourite mug, AI is the shortest route; for an image that has to match an existing object closely, photogrammetry fits; for a true likeness or a part that has to fit, a modeller is needed.

Which photo works

For the AI route: one object, sharp, even light, a calm background and the whole object in frame. A phone photo is fine. More photos of the same object from different angles help, because the generator then has to guess less. What does not work: several objects mixed together, hard shadows, glass and mirroring metal, and photos where part of the object is cut off.

For relief and lithophane the opposite applies: there, contrast and composition count, not the spatial shape. A portrait with a plain background gives a better lithophane than a busy holiday photo. For photogrammetry: walk around, overlap each photo, keep the same light from every side.

From model to print

Once you have a printable file, from whichever route, you upload it, the site measures it and you see the price straight away. We print in resin as standard, one colour per print, at a fixed price per size class based on the height of the print, from €23.95. A model from the AI designer or the studio you then have printed; an STL from another tool you upload at Print my file.

The model carries colour and texture in the file; the print itself is one colour: tinted, or raw grey to paint yourself. If you want colour on the object, you paint it. Full colour comes later; read what we know about it at full colour 3D printing.

Frequently asked questions

Can AI make a 3D model?

Yes. An AI generator turns one to 4 photos or a description into a closed 3D model you can print. The line is nearly right, not a copy: the shape is usually correct, fine details and the back are an estimate, a face does not truly resemble anyone and the model is not dimensionally accurate. If it fails, you get the credit back.

Can it really be done from one photo?

Yes. One photo gives a complete model; the back and sides are then an estimate, which shrinks with two to 4 photos from different angles.

Do I get an STL I can print?

Yes. From the AI designer you download STL, 3MF or OBJ for printing and GLB with colour; relief and lithophane generators deliver an STL. A photogrammetry model often comes as OBJ or GLB and usually still needs cleaning up before it is printable.

Is an AI model dimensionally accurate?

No. The shape and proportions are usually right, but it is not a measurement. For a part that has to fit, have a model designed to dimensions.