Can an AI agent swap the person in my video? Ask Remixy

In a Sume chat, Remixy remakes a clip from a reference and can use H3 Max Recast for a person swap. What to give it, and what it should say before paying.

5 min readSume
All posts

What Remixy is for

Sume's agents chat has specialists. Remixy is the one that remakes a video from a reference you point at. Its description in the product says it uses Seedance 2.5 video-to-video, H3 Max Recast, Genjutsu or reference remix before it starts from a blank prompt.

Person swap is the Recast case. fal's model page, read 2026-10-03, describes Recast as recasting the people in a video with reference photos while preserving the source motion, camera, cuts and audio. Sume exposes that as the h3-max-recast model, 5 to 30 seconds, 768p or 1080p.

What to give it

The agent's instructions in the repository separate the four routes by what the request keeps. Hand over the clip and the photos and say what should stay the same.

  • The source clip, as a file in the chat or a link it can import. Keep it under 30 seconds with no shot over 15 seconds if you want Recast.
  • One photo per new person, up to four. Say which photo replaces which person if the order is not left to right.
  • A plain statement of what stays: "same camera, same cuts, same sound, only the people change".

What it should do before paying

The instructions tell the agent to say the route and why in one line before the first paid call, and to dry-run the request to quote the price. That is the check worth waiting for. For a Recast the quote follows the source length: fal lists $0.30 per second at 768p and $0.45 at 1080p, and Sume bills list with its 1.25 margin. A 12-second clip at 768p is a different number from a 30-second clip at 1080p, and you should see it first.

If Remixy proposes Seedance 2.5 video-to-video instead, that is by design. The same instructions say that route tends to hold a one-take clip best, and that Recast and Genjutsu stay fair picks. Ask for Recast by name if the soundtrack and camera must survive untouched.

When it will say no

Recast has no text-only mode and keeps the source's length, so it cannot change what is said. A new script or a new voice is a different route: a rebuild or a reference remix. For that, the agent keeps the structure and swaps in your product and words instead.

Only swap in people who have agreed to appear. Sume does not check consent for you.

Sources

Related posts

More in Agents

All Agents posts

Written by Sume