Skip to content

DiPIR

Insert virtual objects with recovered lighting

Visit official site

What DiPIR is

The method is a physics-based approach from the Toronto AI lab and NVIDIA Research for realistic object insertion.

What you can do with it

  • Recover lighting from a background image
  • Insert a virtual object into a scene
  • Adjust material and tone mapping

Pros & cons

✓ What we like

  • Indoor and outdoor scenes are covered
  • A diffusion model guides the process

! What to watch out for

  • Lighting comes from a single image
  • Evaluation uses Waymo and indoor panoramas

FAQ

Who developed it?

The Toronto AI lab and NVIDIA Research.

What is recovered?

Scene lighting from one image.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review