How to Turn One Photo Into a Stereoscopic 3D Image

One ordinary photo turned into a stereo pair by AI, left-eye view on the left. Photo: Andy Morffew, CC BY 2.0, via Wikimedia Commons; converted to a stereo pair by Owl3D.

Which route fits you

Last checked October 17, 2026: makers' pages, read from outside the US.

How one photo becomes two views

A depth map is a grey picture of how far away each part of the photo is, usually with near as white and far as black [2]. It gives relative depth, not distances in metres [2].

The software then shifts each pixel sideways by its depth, near things more than far ones, one way for each eye. Where two pixels land on the same spot, the nearer one wins, so the subject covers the background. Behind each edge the shift uncovers background the camera never saw; filling it in causes distortion at those edges [5][6].

If you only need the depth map, for engraving, Blender or ControlNet, use the Depth Map Generator.

Free in your browser: a red-cyan picture

1. Open the Anaglyph Maker and choose "One normal photo". The first time, your browser downloads the depth model (27 to 50 MB) and keeps it [1].
2. Upload your photo. JPEG or PNG, and HEIC where your browser can open it; the photo stays on your device [1].
3. Look through red-cyan glasses, red over the left eye. If near things look far, turn on "Swap left and right" [1].
4. Press "Download PNG". It carries a small "made with owl3d.com" mark [1].

From one photo the tool puts the nearest thing at the screen and everything else behind it, at a fixed, moderate strength [1]. The "Horizontal offset" slider moves the whole scene forward or back; it does not add depth [1].

A side-by-side image for a VR headset or 3D TV

iw3 is a free, open-source program that can "convert any 2D image/video into side-by-side 3D image/video" [5]. On Windows you start it with "Run iw3 GUI.bat"; the first run "may take some time as it needs to download large model files" [5]. Tip: add _LRF_Full_SBS to the file name so VR players such as SKYBOX and DeoVR recognise the layout [5].

The Owl3D converter is a desktop app for Windows and Mac that turns ordinary photos (JPG, PNG) and videos into side-by-side, half side-by-side or top-bottom 3D with AI, on your own computer. It is not instant; it is the route to the best 3D Owl3D makes, and the depth is still an AI estimate. It has a free plan; for video that is 60 seconds per conversion, with no watermark. It does not save anaglyphs: load its side-by-side result into the Anaglyph Maker for that [1].

From Owl3DOwl3D converterA desktop app for Windows and Mac that turns 2D videos and photos into 3D.

Immersity for Web, from Leia Inc., saves 3D images "with parallel, cross-eyed or anaglyph" layouts [7]. More on these tools: 2D-to-3D converters compared.

Parallel or cross-eyed? Parallel puts the left eye's picture on the left, "suitable for most digital viewers"; cross view swaps them, for merging by crossing your eyes [8]. Try it below: cross your eyes until a third eagle appears between the two.

If you write code: open-source depth models

• Depth Anything V2 (2024): the Small model is under the Apache-2.0 licence, Base and Large under CC-BY-NC-4.0, a non-commercial licence [3]. Its makers link an online demo [3].
• Depth Anything 3, released in November 2025, includes models that predict depth from a single image [4].
• ZoeDepth and MiDaS, named in older guides, were archived in 2025 and are read-only [11][12].
• stable-diffusion-webui-depthmap-script turns a generated or your own depth map into side-by-side or anaglyph images, also without Stable Diffusion, with a choice of how gaps are filled [6].
• Google Colab, a hosted notebook service used in a browser, is free of charge, but in the free version "access to expensive resources like GPUs is heavily restricted" [10].

When the 3D looks wrong

What people who tried it report

Users' words on GitHub; we have not tested them.

• Flat foreground. April 2024, on why some conversions pop out more: "The type of depth map used: Depth Anything. Res101 Boost flattens foreground elements while providing detailed backgrounds." [16]
• Gaps in the code route. September 2024, on the depthmap script: "The fill is choppy and makes the stereo image look bad." A project collaborator replied: "Stereoscopic functionality is indeed comparatively primitive." [17]
• Cross-eyed can be easier. February 2023: "I can't parallel view to save my life". A second user, March 2023: "I can do both but parallel is much harder." [18]
• In a headset. November 2023: "The side-by-side images work in my VR viewer (GizmoVR) but the image isn't "immersive."" [19]

Frequently asked questions

Can AI make a real 3D image from one photo?

It makes a convincing stereo image, but the depth is estimated. Two photos taken a few centimetres apart give true depth [1].

Is a "2D to 3D image converter" the same thing?

Usually not. On October 17, 2026, most top US results for that phrase were tools that build a 3D model from a photo. This guide is about stereo pictures, one view for each eye.

How do I view a side-by-side 3D picture?

In a VR player on a headset, on a 3D TV set to side-by-side, or on a glasses-free 3D screen. The Owl3D Shift, with the Owl3D app running on the computer, shows a side-by-side still on its screen in 3D.

I have two photos, an MPO file or a 3D camera picture. What then?

That is a real stereo pair: line it up and save it with the Stereo Photo Maker. For iPhone spatial photos, see spatial photos.