Minimax H3 Recast

It performs cast substitution rather than video generation: the portraits you supply take the place of people already in your footage, while movement, framing, edits and audio all remain from the original.

Between 5 and 30 seconds, with every individual shot under 15 seconds. Split longer takes before uploading, because the result is returned at the same length you supplied.

Up to four, each with its own portrait. The tool maps them to the leads in screen order by default, so arrange uploads accordingly — or change the assignment with a written instruction.

One sharp, well-lit, front-facing portrait per person. Half-body framing carries hair and clothing across more faithfully than a tight head crop. Avoid sunglasses, hats, hidden faces, group photos, heavy filters and blurry images.

Start with a low-resolution pass to confirm the pairing, then render the deliverable at 1080p. Since pricing follows clip length, that first pass is the cheapest way to catch a bad match.

When the footage itself needs to change — different motion, different camera work, a different setting — because all of that comes from your source. Results also weaken with dense crowds, tiny or distant faces, fast movement and objects crossing a face after a hard cut; across long takes, or when someone turns to profile, the likeness can drift back toward the original performer.