Skip to content

InseRF

Insert new 3D objects from text and a 2D box

Visit official site

What InseRF is

The method is described as a way to generate new 3D objects from a text prompt and a 2D bounding box, then insert them into a NeRF-reconstructed 3D scene.

What you can do with it

  • Generate a new 3D object from text and a 2D box
  • Convert a 2D edit into a 3D object model
  • Place and blend the object into the scene

Application scenarios

  • Insert furniture decorations into a 3D indoor scene
  • Add buildings or props to a city scene
  • Place plants or scenery in an outdoor landscape

Pros & cons

✓ What we like

  • No explicit 3D information is needed
  • Insertion is described as 3D consistent

! What to watch out for

  • Input needs a reference viewpoint with a 2D box
  • Evaluation was run on multiple 3D scenes

FAQ

What input does it need?

A text description and a 2D bounding box in one reference viewpoint.

What scenes are named?

The material lists indoor, city and outdoor landscape scenes.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review