A one-tap AI video template is a photo in and a finished clip out
start_template_run needs the template's inputs, usually a photo. It is not a cloned workflow and not a scene-planned movie.
A one-tap AI video template is a photo in and a finished clip out. Attach a photo. Get the template's finished video. start_template_run kicks off one of Versely's one-tap templates after every required input is a hosted URL — usually a photo attached with the chat + button. It is not a scene-planned movie, not a public workflow clone, and not a blank generate. Running it as any of those wastes credits on a template that never received the still it is built to consume, or on a movie you did not need.
Run a one-tap AI video template is the named capability. The templates catalog is where the formats live: finger snap, mugshot, the rest of the one-tap shelf. get_template tells the agent which inputs that template needs. get_template_run and cancel_template_run support the live card. You do not poll status in most cases; the UI subscribes.
What this job actually is
You pick a template by name or id. You attach the photo it asks for. Once every required input is gathered, the agent starts the run and a live-updating card tracks progress. Output is that template's finished video, built from your upload. Each template lists its own estimated credit cost before you run it.
No timeline. No scene drafting. No "write me a three-act structure." The creative choice already happened when someone built the template. Your job is the still, and the decision to use this format.
What wastes credits
A movie brief stuffed into a template name. "Make a short film of my product" is create a multi-scene AI movie. Templates do not take a scene list. They take the inputs on the template.
Cloning a public workflow and calling it one-tap. Recreate a viral template clones a workflow into your account (recreate_public_template). Saved, not auto-run. Different object, different cost shape. One-tap start_template_run is the finger-snap / mugshot class: photo in, clip out, no cloned scene graph.
Starting before the photo is a hosted URL. The tool wants inputs as URLs from the attach button. request_asset_upload is for workflow assets, not for this. Launching a run with a missing still is a failed run you still have to notice.
Using a template as a generic video generator. If no template matches, generate. Forcing a mugshot template to "just be a product ad" is how you pay the template's estimate for a joke format you will not post.
The test
Is there a named one-tap template, and do you have the photo it requires?
If yes, this is the job. Attach, run, watch the card.
If you need a multi-scene story, make a movie. If you need a public workflow's scenes in your account, clone that template. If you need a custom pipeline, spawn a task. This is one named agent job. Using it for a different job wastes credits.
FAQ
Do I pick the video model inside the template?
No. The template already chose its stack. You supply inputs. If you wanted to pick Veo versus Kling per shot, you wanted a generate or a movie, not one-tap.
Can I cancel a run that is in progress?
Yes. cancel_template_run exists for that. Closing the chat is not a cancel.
Is this the same as saving a workflow I can rerun with a new plot?
No. Saved video workflows rewrite a plot around locked characters. One-tap templates do not write plots. They apply a format to your photo. If you want the rerun-with-a-new-plot shape, save and reuse a video workflow.
What if the template wants two photos and I only attached one?
The agent should gather every required input first. Starting with a partial inputs object is how you get an error instead of a video. Attach what get_template lists. Then start.