Field guide · 04
3D scanning: photogrammetry and splats
Photogrammetry, LiDAR and the newer Gaussian splats all have their place. Choose based on speed, fidelity and how you plan to use the asset.
The options compared
| Method | Strengths | Weaknesses | Best for |
|---|---|---|---|
| DSLR photogrammetry | Highest-fidelity meshes and textures | Slow to process, and you cannot confirm it worked until you are home | Hero assets, textures |
| iPad LiDAR scan | Fast, and processes on the device | Lower detail and fidelity | Quick spatial reference, on-the-day checks |
| Drone photogrammetry | Covers buildings and large items cheaply | Weather-dependent, needs a scale reference | Environments, large objects |
| Gaussian splats (drone, phone, SuperSplat) | Fast, very photoreal capture of light and viewpoint | Not a clean mesh, harder to relight or edit, newer workflow | Backgrounds, projections, look development |
| Point clouds | Accurate spatial data | Not render-ready on their own | Measurement, an alignment base |
Tools
- Agisoft Metashape (Standard) for photogrammetry processing (paid).
- Instant Meshes for fast retopology (free).
- fSpy for lining a CG camera up to a still (free).
- Blender for clean-up, retopology, and as the hub everything passes through.
Clean-up is a skill
Scans are never drag-and-drop. You still have to clean the mesh and retopologise, which is a real skill that a lot of VFX artists never pick up. Budget the time, or a specialist, for it.
Gaussian splat cameras and drones are a big change and worth looking into. New drones are arriving that can capture full 360-degree HDRI and splats. Weigh them against photogrammetry though: splats look superb but are not yet a tidy, editable mesh.
Something we are still testing: the best combination of photogrammetry, splat capture and clean-up for a given asset.
Coming soon Walkthrough video
Coming soon Example project files