A toy autostereogram renderer demo in threejs/webgpu, using depth estimation to produce the input depth map.
This is a standard "divergent" magic-eye-style autostereogram, which means you want your eyes to converge "beyond" the image on the screen, while allowing your eyes to focus on the screen.
Normal autostereograms are hard enough for most people to view, but this demo is particularly challenging, so don't be discouraged.
Some viewing tips for this demo:
- Try freezing the image while posing in front of your webcam (use the Controls UI in the top right of the demo screen)
- Try resizing your window to smaller or larger sizes (you may need to unfreeze or refresh the demo after resizing too)
- One method that works for many people: Start with your face right up to the image on the screen, then slowly move your head back until the image converges into focus
Good luck! 🤓
For something that works well on mobile with any still image, try Magic-eye anything.
