Google's Atlas video model nails pixel-perfect camera control incl. fisheye and distortion models
teortaxesTex · x · 2026-09-04
Google's Atlas video generation model precisely follows input camera parameters, including non-planar projections like the Brown-Conrady distortion model and Kannala-Brandt fisheye model. Team member Pan Bhamidipati invented a novel camera conditioning method that reportedly works beautifully; 3D vision practitioners call it a game-changer for the field.
More from Multimodal
- Study: frame selection dominates long-video MLLM accuracy; spatial compression is nearly free if reinvested — Prakhar Khatri · 2026-09-04
- Recraft V4 Styles in practice: keeping brand visuals perfectly consistent — OVolosin82152 · 2026-09-04
- Zero 3D experience needed: GPT-6 Astra builds a Brittany street scene in Blender end-to-end — Dimillian · 2026-09-04
- Synthesia launches Assistant: chat-based AI video creation from idea to finished video in minutes — synthesiaIO · 2026-09-04
- OmniCam brings Blender-like 3D camera control to AI video workflows in ComfyUI — Main_Creme9190 · 2026-09-04
- StreamTalk turns audio into SMPL-X body gestures, rig of your choice — multimodalart · 2026-09-04