GAE generates video and 3D geometry natively in a shared geometry latent space, code released
yshan2u · x · 2026-09-22
A team including researchers from CUHK introduced GAE, a generative model that performs true "3D-native generation" rather than reconstructing geometry from rendered RGB.
- GAE compresses geometry-foundation features into a shared latent space; a conditional flow model generates directly there.
- A single generated latent decodes both RGB appearance and 3D point clouds, aligning appearance and geometry by construction.
- Demos include camera-conditioned world generation and text-to-image pairs, with point clouds generated natively rather than reconstructed.
Code and models are already released, with a project page available.
More from Multimodal
- Ditto launches swap-anything video tool to replace any video's main character — yoobinray · 2026-11-12
- Fable 5.2 impresses with pixel art gaussian splats, posts go viral — Flomerboy · 2026-11-12
- Seedream 5.5 nails a consistent character sticker set at just $0.02 per image — azed_ai · 2026-11-12
- Ant Group's open-source Ming-3.0-tts-0.5B tops Open TTS Leaderboard with 0.88 WER — NielsRogge · 2026-11-12
- Viggle Remake recreated in ComfyUI: swap any character in your videos with MiniMax H3 — TheLocalLab · 2026-11-12
- Pika Integrates Seedream 5.5: Split Images Into 100+ Layers for Precise Multi-Turn Edits — aziz4ai · 2026-11-12