Skip to content

ViewCrafter

High-fidelity novel views from one or sparse images

Visit official site

What ViewCrafter is

The method is described as an approach that combines video diffusion models with point-based 3D cues to synthesise high-fidelity novel views of general scenes.

What you can do with it

  • Synthesise a novel view from one image
  • Control camera pose with coarse 3D cues
  • Extend coverage through iterative view synthesis

Getting started

  • Visit the site and read the project introduction
  • Choose single-view or two-view synthesis
  • Upload a reference image or image set
  • Generate a preview with the video diffusion model

Pros & cons

✓ What we like

  • Camera trajectory planning expands the view range
  • Scene-level text-to-3D generation is listed

! What to watch out for

  • A reference image or image set is required as input
  • The pipeline relies on video diffusion model output

FAQ

What inputs are used?

The material names single or sparse images with point-based 3D cues.

What does it optimise?

The feature list names 3D-GS representation for immersive experience and real-time rendering.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review