Finally, we show more editing results in Figure 9. DreamFusion [ 25 ] introduced Score Distillation Sampling SDS , which facilitates the optimization of differentiable image parameterizations DIP by using diffusion models to compute gradients in the form of:. Experiments on diverse text prompts demonstrate the significant performance of our method. To ensure that we are establishing correspondences only between the foreground objects, i. The total training time amounts to approximately 3 hours.
nest...