Product-drop videos: one bulk queue or a scheduled run per drop?

For a product-drop series, use a bulk Format queue when every drop's inputs exist today and you want them rendered now, and use a schedule when each video should start on a date. A queue takes 1 to 100 items with a window of 1 to 16 and starts immediately; a schedule fires on a 5-field cron expression with a time zone and keeps its instructions between runs.

The two shapes

A bulk queue is one POST with an items array. Each item is the same body as a single Format run, and the queue dispatches them in order within the concurrency window. A schedule is an Agents automation you create in the dashboard: its instructions persist, caller data arrives separately as input, and a cadence or an API call starts it.

A rule of thumb

Ask when the drop date is known. If you have twelve drops with finished briefs and want twelve renders ready for review, a queue is the shape. If you want one render each Thursday from a feed that changes, a schedule is the shape, because it will exist next week with no work from you.

A mixed series works too: a schedule that, on each firing, reads the upcoming drop and renders it, plus a queue for the back catalogue.

import os, requests
items = [{"instruction": f"Drop {i}: {name}",
          "generation_spend_cap_usd": 8}
         for i, name in enumerate(["tote", "shirt", "cap"], 1)]
r = requests.post(
    "https://api.sume.com/v1/formats/acme/product-promo/bulk-runs",
    headers={"Authorization": "Bearer " + os.environ["SUME_API_KEY"],
             "Idempotency-Key": "drops-2026-10"},
    json={"concurrency": 3, "items": items}, timeout=30)
print(r.status_code, r.json()["data"]["status_url"])

Traps on each side

On the queue side, completed means every item is terminal, not that every item succeeded, so read counts.failed. On the schedule side, the default $1.00 cap will stop a video run early unless you set the cap, and the schedule's trigger type cannot be changed after you create it.

Neither shape has a queue-level or schedule-level total budget. Add up the caps yourself before you start.

Limits

A queue is capped at 100 items, so a longer series needs several queues with different idempotency keys. A schedule cannot be created or edited through the Developer API; that happens in the dashboard.