pix2pix3D extends conditional generative models with neural radiance fields to enable explicit 3D user control over image synthesis. Given a 2D label map such as a segmentation or edge map, the model learns to synthesize corresponding images from different viewpoints. It assigns labels to every 3D point in addition to color and density, enabling simultaneous rendering of images and pixel-aligned label maps. The project includes an interactive system for editing label maps from any viewpoint.