Skip to content

MorphCut

Smooth jump cuts and remove filler words

Visit official site

What MorphCut is

The framework is described as a novel approach to smoothing jump cuts in dialogue video, fusing information from other source frames through a mid-level representation driven by DensePose keypoints and facial landmarks. The framework interpolates keypoints and landmarks between end frames around a cut and synthesises pixels with an image translation network.

What you can do with it

  • Remove filler words and repeated words from dialogue video
  • Smooth jump cuts into seamless transition video
  • Handle cuts that remove filler words, pauses or random cuts

How transitions are built

  • Fuse information from other source frames with a mid-level representation
  • Interpolate keypoints and landmarks between end frames around a cut
  • Synthesise pixels from keypoints and source frames with an image translation network
  • Pick the most suitable source for each keypoint through cross-modal attention

Application scenarios

  • Remove unneeded words and jump cuts in editing software for fluent dialogue video
  • Improve video content and viewing experience on an online education platform
  • Remove unneeded pauses and words in advertising production

Pros & cons

✓ What we like

  • Keypoints and landmarks are interpolated between end frames
  • An image translation network synthesises the final pixels

! What to watch out for

  • Transitions rely on DensePose keypoints and facial landmarks
  • Source frames are selected by the cross-modal attention mechanism

FAQ

What does it remove?

The feature list names filler words and repeated words, along with jump cuts.

What drives the transitions?

The material names DensePose keypoints and facial landmarks as the mid-level representation.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review