How to make an AI video with your friends — without uploading anyone's photo

Group AI videos are the most-shared thing people make in Actoria — a heist trailer for the group chat, a fake travel vlog, a birthday roast. They are also the part most tools get wrong, because the honest way to put a real person on screen is to ask them. Here is the whole flow, start to finish.

Why you can't just upload your friend's photo

Most face-swap apps accept any photo. That is convenient until it is your face in someone else's video. Actoria takes the other route: a real face enters the platform only through its owner. You verify yourself with a short liveness check on your phone, and every friend does the same through a personal invite link. Nobody — including you — can add a person who hasn't agreed. It sounds like friction; in practice it takes each friend about two minutes, and it is the reason group videos stay fun for everyone in them. (More on the mechanics in the consent guide.)

Step 1 — Verify yourself

Create your account, open Actors and choose From photo. You'll be asked to pass a liveness check: a live camera session, not a file upload, so a screenshot or a printed photo won't do. Once it passes, you become an actor in your own library, with a consent badge and an expiry date. From now on you can cast yourself in any scene.

Step 2 — Invite your friends

In the Actors tab press Invite an actor. Give the invite a name, pick the access type — time-limited (30 days, revocable) or single use — and send the link however you normally talk to that person. On their side it is a short flow: read what they are agreeing to, pass the same liveness check, done. They don't need to learn the product; they appear in your library as a ready actor.

Step 3 — Build the scene with several people

Create a project, pick a format (16:9 for YouTube, 9:16 for Stories and Reels) and a style. Add a location — describe it or pick one from your library — then add actors. The frame is composed around the first actor; the others are inserted into the frame one by one, so a four-person shot is built deliberately instead of hoping the model guesses the group right. Iterate on the still until everyone is where you want them: this is the cheap part, so retake freely.

Step 4 — Animate and extend

Describe the action and the camera in a sentence or two (“they walk toward the camera laughing, slow push-in”), choose a model — Seedance 2.5 if you want the clip generated with its own sound — and generate. If the first seconds work but the scene should continue, extend it from the last frame: the next clip starts exactly where this one ended, with the same faces.

Ideas that work for groups

• A movie trailer about your friend group, complete with a dramatic voiceover line in the prompt
• The travel vlog from the trip you never took
• A fake commercial for an inside joke
• A birthday video where the whole crew shows up in a place the birthday person loves
• The band that never existed — one shot per “member”, cut together

If you want more prompts, the AI videos with friends page has the use cases people actually make, and AI video of me and my friends covers two-to-four-person scenes in more detail.

Common mistakes

• Animating before the still is right — fix composition in the frame, not in the video
• Packing six people into a close-up; wide shots hold groups, close-ups hold pairs
• Forgetting consent expiry mid-project
• Writing the whole movie into one prompt — one action per scene, then extend