Skip to content

MVDrag3D

Drag-edit 3D shapes across multiple rendered views

Visit official site

What MVDrag3D is

The framework is described as a 3D editing method that applies consistent drag edits across several rendered views. Generation and reconstruction priors keep the views aligned.

What you can do with it

  • Prepare or upload a 3D model
  • Drag parts of the shape across views
  • Improve geometric alignment with the deformation network

Reference cases

  • Car body lines are reshaped by a designer
  • Film 3D scenes are adjusted by a visual effects user
  • Building model details are refined by an architect

Pros & cons

✓ What we like

  • A multi-view scoring function improves view consistency
  • Several object categories and 3D representations are named

! What to watch out for

  • Edits begin from a prepared or uploaded 3D model
  • View alignment depends on the deformation network

FAQ

What guides the drag edit?

A multi-view diffusion model is used as the generative prior.

What is rebuilt afterwards?

The edited object is reconstructed as 3D Gaussians.

Last reviewed: 2026-11-15

More AI design tools tools

View all →

How we review