Skip to content

SVFR

A unified framework for video face restoration

Visit official site

What SVFR is

The framework is described as a unified approach to video face restoration that joins restoration, colourisation and inpainting with video diffusion priors.

What you can do with it

  • Combine BFR, colourisation and inpainting
  • Use learnable task embeddings
  • Share features across sub-tasks
  • Visualise results on low-quality face video

Getting started

  • Download the code and model files
  • Choose a task combination
  • Preprocess the low-quality face video

Pros & cons

✓ What we like

  • Tasks are combined in one framework
  • Temporal coherence is addressed

! What to watch out for

  • It is built around the SVD priors
  • Task combinations are chosen by the user

FAQ

Which tasks are combined?

The material lists blind face restoration, colourisation and inpainting.

What prior is used?

The material names Stable Video Diffusion generation and motion priors.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review