Skip to main content
Magejutsu edits a video instead of generating one. Give it a clip you already have and photos of who or what should be in it, and it rebuilds the cast and the scene while the performances, camera moves and timing stay as filmed. A director model watches your clip first and writes the edit for you.

Overview

What it supports on Mage

Best for

  • Putting yourself, a character, or a client into an existing clip
  • Changing a location, outfit, or product without reshooting
  • Keeping a real performance, dance, or camera move intact
  • Restyling footage you already own

An example prompt to start from

Map each reference to its subject by position: “image 1 is the person on the left”. One clear, front-lit photo per person is enough, and you can send up to 30 references: characters, outfits, props and places.

Models

Modes

Magejutsu prepares the source for the mode itself; there is nothing else to set. Directions are optional: without them, the director maps your references to the clip on its own.

Limitations

  • Sources run 4 to 30 seconds, and the result matches the source length and shape.
  • It edits footage rather than inventing it, so a source clip is required.
  • The result is billed in Gems on every plan.

Use via API

Magejutsu is available through the Mage API as the magejutsu architecture. Its fields, tokens, defaults, and price are on its API reference page.
Last modified on October 31, 2026