wedding-immortalist

3D Gaussian Splatting reconstructs venue-scale memory spaces from wedding photos and footage. The pipeline uses COLMAP-based structure-from-motion to build geometry, trains 3DGS models with wedding-optimized settings, and outputs an interactive viewer for immersive playback across ceremony, reception, and venue spaces.

Face clustering groups guests from wedding photos using facial embeddings, then aesthetic scoring selects top photos per person. The result is a curated guest roster with AI-selected best moments for each attendee.

Theme extraction identifies design motifs from your wedding—disco, rustic, modern, or queer celebration aesthetics—and applies them to the 3D memory space. This ensures the immersive environment reflects your wedding's visual identity.

Theatre-mode replay lets you relive ceremony and reception moments in place within the reconstructed 3D venue. The interactive viewer enables spatial exploration of tagged moments and guest interactions across the memory space.

Ingest thousands of wedding photos and hours of footage. The pipeline processes raw images and video to generate COLMAP poses, trains 3DGS models, extracts facial embeddings, and outputs organized memory spaces ready for interactive playback.

No specialized hardware is required beyond standard compute. The pipeline handles COLMAP structure-from-motion and 3DGS training internally; you provide photos and video, and the system outputs a viewer-ready 3D memory space.