This is great – thanks for explaining it all, Mike! In the research paper, the authors compare GS3D to NeRFs, but the scenes train more quickly (perhaps 30 min

This is great – thanks for explaining it all, Mike! In the research paper, the authors compare GS3D to NeRFs, but the scenes train more quickly (perhaps 30 minutes for a scene), and render in real-time. So it's really been a huge upgrade from NeRFs and photogrammetry. I think there's a lot of potential to use GS3D for AR/VR avatars, and also stable diffusion-type scene generation, (where a more-complete scene is generated from one or two photos, or solely imagined by AI). Excited to see where it goes!