DailyPapers (@HuggingPapers)
2025-12-26 | โค๏ธ 309 | ๐ 42
Spatia: Video Generation with Updatable Spatial Memory
Microsoft Research proposes maintaining a 3D scene point cloud as persistent spatial memory to enable long-horizon, spatially consistent video generation with explicit camera control and 3D-aware editing. https://x.com/HuggingPapers/status/2004472492019065042/video/1
๐ ์๋ณธ ๋งํฌ
๋ฏธ๋์ด
๐ Related
- you-could-direct-a-video-like-a-real-3d-world-researchers
- you-could-turn-any-2d-video-into-high-quality-3d-in-a
- you-cant-3d-reconstruct-glass-from-images-wrong-thanks-for
- diffusion-knows-transparency-repurposing-video-diffusion
- what-if-you-could-make-a-robots-ai-brain-more-stable-and