Splatcatcher: LiDAR Capture

Splatcatcher captures everything you need for 3D Gaussian Splatting training in a single session — images, camera poses, and a dense colored point cloud — using your iPhone's LiDAR sensor.

IMPORTANT: Capturing requires the LiDAR scanner, which is only available on iPhone Pro models (iPhone 12 Pro and later Pro / Pro Max). Non-Pro iPhones do not have LiDAR and cannot capture.

Walk around any object or scene and Splatcatcher handles the rest. Each capture records the camera frame, full intrinsic and extrinsic parameters, and accumulates a spatially-deduplicated point cloud from the LiDAR depth buffer in real time. When you're done, export a ready-to-train COLMAP dataset — cameras.bin, images.bin, a colored points3D.bin, and your images — without ever running COLMAP or any other SfM pipeline on a desktop. Just load the folder into LichtFeld Studio (free), Postshot (paid), or any other COLMAP-based trainer and you're good to go.

Or skip the copying altogether: pair Splatcatcher with Forge, our free companion app for Mac and Windows, and tap Send to PC. Your capture goes straight to your own computer over your network — never through our servers — Forge trains the splat with Spirula Studio or LichtFeld Studio, and the finished splat comes back to your iPhone to view. Get Forge at splatcatcher.warp-gate.com/forge.

Features:
- Real-time dense point cloud accumulation from LiDAR depth with spatial deduplication
- Manual shutter mode with volume button trigger
- Continuous auto-capture mode with distance/angle thresholds and blur rejection
- Live camera-perspective and overhead point cloud preview during capture
- Tracking quality indicator with motion and feature warnings
- Loop closure guidance to improve reconstruction accuracy
- On-device pose refinement — an optional one-tap optimization pass tightens camera poses and re-aligns the point cloud for sharper training results, no desktop SfM needed
- Exports a standard COLMAP dataset with per-frame camera intrinsics and extrinsics, so autofocus shifts are carried through to training instead of averaged away
- Colored COLMAP point cloud (points3D.bin), written in the same world frame as the cameras for splat initialization
- One-tap export to Dropbox, AirDrop, Files, etc for transport to PC for processing
- Train on your own Mac or PC: pair with the free Forge app by scanning a QR code, send a capture, and follow training from your phone. Close the app once the upload finishes; training carries on
- Built-in splat viewer: orbit, pan and zoom around your trained splats on iPhone (iOS 27 or later; earlier versions show a preview image)

Output is directly compatible with LichtFeld Studio, Postshot, and other trainers that read the standard COLMAP images + sparse/0 format.

Requires an iPhone Pro model with the LiDAR scanner (iPhone 12 Pro or later Pro / Pro Max). Non-Pro iPhones are not supported for capture.