Skip to content

Follow-Your-Canvas

Extend video to high resolution with diffusion models

Visit official site

What Follow-Your-Canvas is

The technique is described as a diffusion-based video outpainting method that generates high-resolution content. It distributes processing over spatial windows to work around GPU memory limits.

What you can do with it

  • Raise a video from 512x512 up to 1152x2048
  • Keep spatial and temporal consistency with the source
  • Generate richer content around the original frame

Processing steps

  • Prepare a source video that meets the needs
  • Choose a spatial window size for the available GPU
  • Encode the source with the layout encoder
  • Compute relative region embeddings for new areas

Pros & cons

✓ What we like

  • Rich and varied video content is generated
  • GPU memory limits are addressed by window merging

! What to watch out for

  • A source video file has to be prepared first
  • Window size is chosen to fit GPU memory

FAQ

What resolution can it reach?

The features state an example from 512x512 to 1152x2048.

How is memory handled?

Tasks are distributed over spatial windows and the results merged.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review