Robbyant Team Open-Sources LingBot-Map, a Feed-Forward Model That Reconstructs 3D Scenes From a Single Video at 20 FPS

A China-based research group, the Robbyant Team, has released LingBot-Map, an open-source 3D foundation model that rebuilds entire real-world scenes in three dimensions from a single RGB video, running at roughly 20 frames per second on one GPU. The release, licensed under Apache-2.0 and distributed via GitHub, Hugging Face and ModelScope, is drawing attention for combining real-time performance with long-sequence stability that has traditionally been hard to achieve without specialized hardware.