Report
Generative Cinematographer uses 3D camera and subject-motion controls to generate video from one image
The researchers say artists can set a camera path and move selected parts of a subject before generating a shot.
TLDR
The researchers behind Generative Cinematographer describe turning one image into an editable 3D scene so artists can plan camera and subject motion together. Their paper reports improved geometric consistency when viewpoints change. The project page notes that incomplete geometry and depth errors can cause artifacts during large viewpoint changes.
Combined views
21
1 Source, first seen ago
