DreamFusion

by Google Research · open-source · Last verified 2026-03-17

DreamFusion is Google Research's landmark text-to-3D method that uses Score Distillation Sampling (SDS) to optimize a NeRF using a pretrained 2D diffusion model as a supervisory signal, enabling text-driven 3D generation without any 3D training data. While its outputs are slower and noisier than later supervised approaches, DreamFusion's SDS technique became the foundational algorithm for a generation of 3D generation research.