BlenderNeRF

Why we included this project

Getting camera poses that match rendered frames is usually the slowest part of building a NeRF or Gaussian Splatting training set, and this Blender add-on automates that whole step. You arrange a scene and camera path in the viewport, click once, and it exports the renderings together with a transforms JSON carrying the intrinsics and extrinsics, in formats that feed directly into Instant NGP and the standard NeRF convention, plus an optional points3d.ply to seed Gaussian Splatting. Three capture modes cover the usual research setups: frames sampled from a single animated camera, a train/test split across two cameras, and views sampled uniformly on a sphere around the object. That makes it handy for researchers, VFX artists, and hobbyists who want controlled, reproducible synthetic data instead of tracking down real captures. It is a plugin, so it expects you to be comfortable building scenes in Blender rather than being a standalone training tool.