Skip to content

Slicedit

Edit video with text prompts and spatial-temporal slices

Visit official site

What Slicedit is

The technique is described as a zero-shot video editing method that uses a text-to-image diffusion model with spatial-temporal slices to improve temporal consistency.

What you can do with it

  • Process spatial and spatial-temporal slices with a pre-trained diffusion model
  • Extract noise volumes and attention maps through DDPM inversion
  • Inject extended attention maps at specific time steps
  • Combine two predicted noise volumes into the final one

Example uses

  • A person jumping becomes a humanoid robot jumping
  • A cat sunning in the grass becomes a cheetah in the grass
  • A herd of cows grazing becomes a herd of sheep grazing

Pros & cons

✓ What we like

  • The source structure and motion are preserved
  • Inversion and sampling run under source and target text prompts

! What to watch out for

  • Editing depends on a pre-trained text-to-image diffusion model
  • Noise volumes are combined from two predictions to form the result

FAQ

What does it use for consistency?

The material names spatial-temporal slices and extended attention maps injected at certain time steps.

What stays unchanged?

The material states the structure and motion of the original video are kept.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review