MetaView: Generate Explorable 3D Scenes from a Single Image

multimodalart · x · 2026-07-17

Kuaishou introduced the MetaView model, which generates geometrically correct novel camera views from a single image, maintaining accuracy even under significant viewpoint changes. The model is currently available to try on Hugging Face Spaces.

Related event: Kuaishou Unveils MetaView for Single-Image 3D Synthesis(2 posts)→

Original post →

More from Multimodal

Multimodal channel →