Comment by dmarcos

I believe so. This is not a model that generates pixels frame by frame from user input like genie 3. Instead, there’s an actual 3D scene / structure generated (point cloud, 3dgs) from the input images.