Filming takes, and which model to pick
Each script beat films as a take. Kling 3 Omni is the safe default; Veo is stricter about faces.
A take is one script beat filmed as a clip of the creator talking. The Takes step has one row per beat with its own Film button and price.
- Kling 3 Omni (default): the most natural UGC delivery, accepts any realistic creator photo, opens on the composed shot, 60 credits per take at 720p. The right choice for most ads, and what a whole-ad one-shot or a scene films on unless you pick another.
- Seedance 2.5: the most cinematic, priced per 5 seconds, up to 30 seconds per take. Accepts realistic creator photos too.
- Veo: 8-second ceiling, and strict about faces: its face filter refuses photos of real-looking people, so it films from a description of the creator instead; the face is close but can drift between takes.
Takes are as long as their words need. The Script step shows "~Ns to say" per beat and warns when a beat will not fit the engine's ceiling. Trim the line or split it.
Under How it is filmed, choose the model's own voice or the creator's locked ElevenLabs voice. Native takes can add Match the voice afterwards.
A whole ad can film as one continuous shot; longer scripts split into 15-second parts filmed back to back with continuity between them.
Takes run through a frame and transcript check on our servers, with one free re-roll if the words or the product came out wrong.