Sume Formats that return a product still instead of a video

Nine Sume catalog Formats are registered with image output, so a call to one returns a still image instead of a video. They are sume-water-splash-hero, sume-model-product-portrait, sume-magazine-cover-campaign, sume-formula-texture-hero, sume-sunscreen-splash, sume-cream-squeeze, sume-serum-drip, sume-toner-pour and sume-editorial-product-set. Use them when you need a hero still or a thumbnail, not a clip.

What the examples show

Each Format carries an example brief in the Sume repo. The ones below are the ones with a plain subject, and they all ask for a 9:16 image generated from an attached packshot.

Confirm the profile before you rely on it

The catalog docs say to read a Format with GET /v1/formats/sume/{slug} before calling it. The io field returns output_kind, which is image for these. Formats saved before registration have io: null, so a filter on output_kind is a way to make the pipeline fail early instead of paying for the wrong thing.

Why stills are a good first step for a batch

A batch of stills can pick which products deserve motion before you start any video run. Queue the stills in a bulk run, read each child receipt, and move only the winners to a video Format.

The run is still a normal agent turn, with a cap and a receipt. Bind an output_schema with a SumeMediaFile and a primary_output_key so the image URL comes back under a stable name.

The names also help organize a batch. A beauty line with a serum, a toner and a sunscreen can be one queue with an item per product and a Format per queue, so sume-serum-drip, sume-toner-pour and sume-sunscreen-splash each get a queue. The bulk endpoint is per Format, so mixing Formats means separate queues.

Use of the result

The still belongs to the key that made the call, like the media from any Format run. Treat the generated image as a draft to review. Packaging text and labels are the usual place where a generated image drifts, so compare the still against your packshot before it goes in an ad.