Skip to content

Stable Virtual Camera

Render new scene views from a target camera path

Visit official site

What Stable Virtual Camera is

The model is described as a 1.3B-parameter general diffusion model developed by Stability AI for novel view synthesis. It is a Transformer image-to-video model that builds 3D-consistent new scene views from input views and a target camera without extra NeRF distillation.

What you can do with it

  • Generate 3D-consistent new scene views from input views
  • Specify the target camera trajectory freely
  • Produce large viewpoint changes with temporally smooth samples
  • Create seamless looping videos up to half a minute long

Example uses

  • A researcher adjusts the camera path to examine 3D consistency
  • An artist uses different scene views as inspiration for digital painting
  • A teacher generates building views from several angles for a lesson video

Getting started

  • Visit the project GitHub repository for code and documentation
  • Prepare the runtime environment and install the required dependencies
  • Collect input view data in the format the model expects
  • Decide the target camera trajectory, viewpoint and motion path

Pros & cons

✓ What we like

  • No extra NeRF distillation is needed to keep consistency
  • The target camera trajectory can be set freely

! What to watch out for

  • The material states research and non-commercial use only
  • Generation starts from input views and a target camera

FAQ

How large is the model?

The card describes it as a 1.3B-parameter diffusion model.

What is needed to generate a new view?

The material names input views and a target camera, with the camera trajectory set by the user.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review