How recast works.

No whitepaper, no onboarding wizard. Here's the plain-English version of what recast does and how to get the most out of it.

The basics

You give recast a source video (or pick one of the templates) and a few photos of your AI influencer.recast sends that to Higgsfield's Genjutsu model, which recasts the video around your character — same motion, same camera, same lip sync, new face. You get a finished clip back, usually in a few minutes.

There are two modes:

1. Motion transfer

   Keeps the whole performance — movement, camera, lip sync — and rebuilds it around your influencer. Best for dances, talking-head clips, walk-ups.

2. Object swap

   Swaps one thing in the shot — a person, a product, an item someone's holding — and leaves the rest of the footage alone. Best for UGC-style product ads.

Characters & reference photos

Add up to 8 reference photos per render. More angles of the same face usually means a stronger match — it's not magic, the model just has more to work with. Save a set of photos as an influencer on the influencers page so you don't have to re-upload every time, and type @ in the instructions box on the create page to pull a saved influencer's photos into whatever you're working on.

There's also an optional environment photo slot — a reference for the background or setting you want the scene in. It's not a separate parameter under the hood, it just gets added to your reference photos with a note telling the model to use it as the backdrop.

Pricing, in short

You pay per second of output, by resolution (480p / 720p / 1080p). The exact price is set once we read your source video's real length — the estimate you see while picking a video can shift slightly once that happens. Clips run 4–30 seconds; longer uploads get trimmed to fit.

Paying with $CAST gets a small discount over other payment options. If a render fails outright (blocked by the content filter, service hiccup, etc.) the cost goes back to your account as credit — it's automatic, you don't have to ask.

Templates

Templates on the create page are ready-made source clips — pick one instead of uploading your own video. They're either owned by recast, AI-generated, or properly licensed. If you've got a clip you think would make a good template, that's a conversation for the team, not a self-serve upload.

The $CAST flywheel

Renders cost money to run. The recast model routes render revenue toward buying and burning $CAST, so usage — not hype — is what moves the token. You can watch it happen in real numbers on the stats page. $CAST is still pre-graduation and this is all early — no promises about price, timeline, or anything financial. Use the app because you like making videos, not because of what the token might do.

Keep it honest

Every render requires you to confirm you'll post it with an AI-generated label. That's not a formality — people should know when they're looking at generated content. Full rules are in the terms.

Questions?

This is a small, community-driven project — if something's confusing or broken, say so on X. That's where things get fixed fastest right now.