An events team wants a photorealistic, explorable model of a venue: something a client can walk through in a browser to plan a production, not a survey-grade point cloud. The priority is visual fidelity and atmosphere: lighting, materials, and the feel of the space.
That is exactly where 3D Gaussian Splatting fits. Rather than a clean geometric mesh, 3DGS reconstructs a soft, lifelike scene from captured views, ideal for venues, sets, and spaces where the experience matters more than millimetre accuracy.
Photoreal where mesh falls short
Traditional meshes struggle with reflective, transparent, and fine-detailed surfaces: stage lighting, glass, foliage, fabric. 3DGS handles these gracefully, producing scenes that look convincing from any angle and load smoothly in a browser or XR headset. For a venue, that means a walkthrough a client can actually explore, not a technical deliverable that needs interpreting.
- Photorealistic, explorable scenes
- Handles reflective and complex surfaces well
- Browser and XR-friendly delivery
- Great for venues, sets, and marketing
From walkthrough to scene
The capture itself is a four-step rhythm, and none of it depends on Volumet beyond the rental and optional onboarding:
- Plan. Identify hero viewpoints and the path a viewer should take; plan even coverage and consistent lighting.
- Capture. Walk the venue with rented XGRIDS equipment: PortalCam for dedicated 3DGS, or a K2 capture feeding the pipeline.
- Process. Reconstruct the 3DGS scene through the XGRIDS / LCC CyberColor workflow.
- Deliver. Publish a browser-viewable scene or export for XR and virtual production.
What the venue scene delivers
Typical outputs from a 3DGS venue capture like this, subject to device, software, and processing workflow:
- 3DGS scene: a photorealistic Gaussian Splatting scene, explorable in a browser or XR, and usable for virtual production, previsualisation, and marketing walkthroughs.
- Point cloud: an underlying point cloud where spatial reference or measurement is also needed.
- Digital twin seed: a visual foundation for digital twin and previsualisation experiments.