Small business holiday ad creative with AI: one Format, one bulk run
Google showed three creatives building local campaigns with Flow. A shop can run a first-party Sume Format per product and queue up to 100 runs in a bulk call.

A local shop can make holiday ad creative by calling a first-party Sume Format at the reserved sume handle, one call per product, and by queuing a short list as a bulk run of up to 100 Format runs. No fork or install is needed first.
Google's blog post, read 2026-10-01, describes The Small Brief: three creative leaders given unlimited access to Google Flow to build campaigns for local businesses and nonprofits. Sume facts are from the Format catalog, bulk runs and create a run docs.
What did Google's Small Brief show?
The post says Google launched the project in May, with its agency, to celebrate Small Business Week. Three creatives built campaigns for local businesses using Google Flow, and the post links behind-the-scenes videos. It is a showcase of what small teams can make; it gives no cost or performance figures that this post relies on.
Which Sume Format should a shop start with?
The catalog lists the callable slugs today, including sume-product-usage-demo, sume-product-commercial, sume-before-after and sume-slideshow. Read one first with GET /v1/formats/sume/{slug}, which returns its description and an io profile: input_kind is url, text, image or product, and output_kind is video, image or text. A slug not on the list answers 404 format_not_found.
Formats by Sume are called at POST /v1/formats/sume/{slug}/runs, and any key carrying formats:write may call one. The run, its media and its spend belong to the calling key.
How do I queue the whole holiday list?
The bulk route takes concurrency (1 to 16) and items (1 to 100), each item the same body as a single run. It returns 202 with a queue; poll GET /v1/format-run-queues/{id}. Use a fresh Idempotency-Key per batch, because replaying a spent key returns the old queue.
| Item | Value |
|---|---|
| Items per queue | 1 to 100 |
concurrency | 1 to 16 |
| Queue webhook | None; webhooks are per item |
completed status | Every item is terminal, not all succeeded; check counts.failed |
curl -sS -X POST "https://api.sume.com/v1/formats/sume/sume-product-usage-demo/bulk-runs" \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{
"concurrency": 2,
"items": [
{ "instruction": "Holiday gift angle", "input": { "url": "https://example.com/p/1" } },
{ "instruction": "Holiday gift angle", "input": { "url": "https://example.com/p/2" } }
]
}'What should I check before running it?
Read the io profile so input matches, set a per-run generation_spend_cap_usd, and run one item before the batch. For a phase plan across the season, see holiday ad creative by phase.
Sources
Related posts
More in Use cases
- How many ad variants per product? Google says test gradually
Google's YouTube ads guidance says test gradually and favor quality over quantity. A starting plan: 3 variants per SKU in one bulk run, reviewed by a person.
- Change a garment's colour in a photo with a mask_url edit
Recolour clothing with a mask: send the photo as an input reference, a public HTTPS mask_url, and a colour prompt to ChatGPT Image 2.5 on POST /v1/images.
- Overlay a graphic on avatar video: Sume compose overlay step
HeyGen's demo fades a HyperFrames composition over a live avatar. On Sume, put one still over finished avatar footage with timeline compose operation overlay.
- HeyGen avatar new outfit with reference images vs Sume
HeyGen prompt avatars take avatar_id plus up to three reference_images for a new outfit. Sume's photo input takes one image_url per avatar.
Written by Sume