Depth3D vs DepthWeaver

Depth3D converts local MP4 and MOV videos into stereo side-by-side 3D video using on-device AI depth estimation. Import a video, and Depth3D processes it entirely on your device to generate: • A depth preview video • A stereo side-by-side (SBS) video for compatible 3D viewers and displays • Raw depth data (.npy) for advanced workflows After depth data is generated once, you can adjust stereo strength or output resolution and regenerate the stereo video without recomputing depth. All processing happens on-device. No uploads. No servers. No account required.

DepthWeaver turns real-world depth into hidden-image stereograms. Capture a scene with the LiDAR scanner, generate depth from any photo with on-device AI, scan a real object with Apple's Object Capture, or load a 3D model — then watch a hidden image rise out of the noise as you cross your eyes. CAPTURE DEPTH FIVE WAYS • LiDAR live capture on iPhone Pro and iPad Pro — point, freeze, done • On-device AI depth from any photo (Depth Anything V2) • Scan real objects with Apple's Object Capture (Pro devices) — guided orbits turn a real object into a .usdz you can orbit and frame • 3D model import — .usdz, .usd, .obj, .scn with an orbit camera so you choose the angle • Bundled height map and 3D model presets to play with right away PICK A PATTERN • Six procedural generators: random dot, stars, Perlin noise, Worley noise, Voronoi cells, reaction-diffusion • Import any image from your photo library as a custom tile • Built-in texture library TUNE EVERY KNOB • DPI, depth strength, separation, oversampling • Remap input and output depth ranges to push the subject forward or back • Denoise quantized 8-bit imports for clean, banding-free output • Invert depth in one tap • Live preview redraws as you slide PRIVATE BY DESIGN Everything runs on device. No account, no upload, no tracking, no ads. Your photos and 3D scans never leave your device. UNDER THE HOOD Parallelized implementation of the Thimbleby–Inglis–Witten autostereogram algorithm with W. A. Steer's link-based hidden-surface refinement. Powered by ARKit, Object Capture, Core ML, SceneKit, Metal, and Core Image. Output is upscaled so the minimum dimension is at least 960 px — print-ready or share-ready. SHARE Export high-resolution PNGs to Photos, Messages, AirDrop, or any app with a share sheet. Available in English and French. For iPhone and iPad (iOS 18 or later) and Mac with Apple silicon (macOS 15 or later). LiDAR capture and Object Capture scans are iPhone and iPad only. Capture a scene. Pick a pattern. Stare into it.