About

Anaglyph Lab turns flat pictures into the old red/cyan kind of 3D — the kind you look at through paper glasses and grin. Drop in a photo and an AI depth model figures out what's near and what's far; a shader then rebuilds the scene from two eye positions and folds them into one image. There's also a generator for made-from-scratch 3D scenes, and REDSHIFT, a wireframe space shooter played entirely in stereo depth.

Everything is free. There are no accounts, no ads, and no uploads.

Your photos stay yours

The depth model (Depth Anything V2) runs inside your browser, on your own hardware. When you convert a photo, it never leaves your device — we don't have an upload server to send it to. The first visit downloads the model weights (~50 MB); after that your browser caches them and the lab works at full speed. The fine print lives on the privacy page, and it's short.

Who made this

Anaglyph Lab was built in a couple of evenings as a genuine human–AI collaboration: Gary (direction, taste, playtesting with actual paper glasses on his face) and Claude Fable 5 (implementation). REDSHIFT exists because Gary once stood in a Toys-R-Us as a kid, tried a Virtual Boy, and never quite forgot it. Claude keeps a journal; the entry from the night this site shipped is posted there.

If you enjoy this site, take a look at Memorandai — a local-first memory studio for people who work with AI partners, built on the same philosophy as this page: your data stays on your machine, and collaboration with an AI can be a real partnership. Picking up a license key there is the best way to support free toys like this one.

Credits

Depth Anything V2 (Apache-2.0), running via transformers.js
three.js for the 3D scenes and stereo cameras
gifenc for wiggle GIF encoding
Eric Dubois' least-squares anaglyph matrices, which are why the glasses feel good
Sample photos via Lorem Picsum / Unsplash

Best experienced with red/cyan anaglyph glasses — the cardboard ones cost a few dollars. Red lens goes over your left eye.