Red Carpet is a one-tap file, not a prompt you rewrite weekly
/templates is a photo in, a clip out. A new generate is a different object.
/templates is a photo in, a clip out. A new generate is a different object.
Red Carpet puts you (and a plus-one) on a film-festival carpet — tuxedo, sequined gown, paparazzi flashes. Inputs: Person 1 (left) and Person 2 (right), both optional, both with defaults. Estimated cost on the card: about 68 credits (135 at the listed full rate). You do not rewrite a Veo prompt every Friday to chase the same gag. You attach the stills the template already knows how to consume. You run it.
A blank generate — "make me on a red carpet, cinematic, crowd, cameras" — is a different SKU. Different bill. Different failure mode. The template already chose the stack. Your job is the photo.
Photo in. Finished clip out.
Run a one-tap AI video template is the named agent job. get_template lists the inputs. start_template_run starts when every required input is a hosted URL, usually a photo attached with the chat + button. A live card tracks the run. get_template_run and cancel_template_run support that card. You do not poll status in most cases. Output is that template's finished video.
Red Carpet's two image fields are optional. Defaults exist so the format still runs. If the gag is you, attach you. If the gag is a couple, attach both. Starting with a partial inputs object is how you get an error instead of a clip. Attach what the template lists. Then start.
No timeline. No scene list. No "write a three-act arrival." The creative choice happened when the template was built. Weekly novelty is a new photo, not a new paragraph.
A weekly prompt is not this template
People open a text-to-video row and paste last week's carpet prompt with a new adjective. That is a generate. It will invent a face that is not the upload, a crowd that does not match, a gown the still never wore. It will also not be the 68-credit one-tap object. You paid a model to approximate a template you already have.
Forcing Red Carpet to be a product ad, a talking brand film, or a multi-scene movie is the other miss. Templates do not take a plot. Create a multi-scene AI movie is a different named job. Cloning a public workflow (recreate_public_template) is a different object again: saved, not auto-run. One-tap is finger-snap / mugshot / red-carpet class: photo in, clip out, no cloned scene graph.
If no template matches, generate. Do not run Red Carpet "just to have motion." You will pay the template estimate for a format you will not post.
Do not pick the model inside the carpet
The template already chose its stack. Asking for Veo versus Kling on a one-tap run is how you discover that this job does not expose that switch. If you needed to pick a row per shot, you wanted a generate or a movie, not Red Carpet.
Cost is the number on the template card, shown before you run. It is not a per-second cinematic meter you can talk down in the prompt. Cancel an in-flight run with cancel_template_run. Closing the chat is not a cancel.
Use it when the post is the carpet gag. Skip it when the post is a product, a talker, or a story. Rewrite nothing weekly except the still.
FAQ
Do I have to upload two photos?
No. Both person fields are optional and ship with defaults. Attach one still for a solo arrival, two for a plus-one, or none to see the format. The template still runs.
Can I change the gown or the crowd in the prompt?
Not as a one-tap. The format is fixed. A new generate with a rewritten prompt is a different object — different cost, different face risk, no template card.
Is this the same as saving a workflow I rerun with a new plot?
No. Saved video workflows rewrite a plot around locked characters. Red Carpet does not write plots. It applies a format to your photo(s).
What if I want a different model for the same joke?
Then you are not in this template. Generate, and accept that you are no longer in the one-tap, ~68-credit, photo-in-clip-out path. Pick that on purpose.