liveBook · Manning

Video-to-video synthesis

As explained in Generative AI in Action

Video-to-video synthesis is the generation of a new video from a source video while preserving aspects of its underlying structure.

Video-to-video synthesis generates a new video from a source video by preserving its underlying structure while changing its content. It supports video-generation workflows where motion or scene organization must remain recognizable across a transformation. section 5.4

Distinguish it from neighboring video-generation methods:

  • Text-to-video synthesis creates video from a written prompt. section 5.4
  • Image-to-video synthesis creates video from a source image. section 5.4
  • Video-to-video synthesis starts with an existing video and retains its structure while producing different content. section 5.4

How Generative AI in Action covers it

How deep it goes
Defines structure preservation and learned transformation but stops short of implementation details, architecture, and mathematical derivation.
Examples
Uses a source video to create another video with the same structure but different content.
What you can do after
Evaluate whether a video-generation workflow is conditioned on text, an image, or an existing video's structure.
How it connects
Connects video generation with text-to-video synthesis, image-to-video synthesis, and GAN-based video generation.
Cover of Generative AI in Action
liveBookmanning.com
Read Generative AI in Action on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Generative AI in Action. The section links point to the parts of the book it draws on.