Asset Studio 1-Click A/B Testing vs a Sume bulk run of hook variants

Google's Asset Studio update adds Gemini Omni video creation and 1-Click A/B Testing inside Google Ads. If the test lives only in Google, use that. If you want the same five hooks tested on TikTok, Meta and Google, make the variants once outside any one console: a Sume bulk run queues up to 100 Format runs, up to 16 at a time, and returns one clip per item.

What did Google announce?

Google's post is dated May 20, 2026 and was read on 2026-10-01.

How do I queue hook variants as a bulk run?

Each item is the same body as a single Format run, so a variant is one instruction with a different hook line. Mint a fresh Idempotency-Key per batch: replaying a spent key returns 202 with the old queue. Put a per-item communication.webhook_url on every item if you want a callback, because the queue itself has no webhook.

curl -X POST https://api.sume.com/v1/formats/sume/sume-video-hook/bulk-runs \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: q4-hooks-batch-001" \
  -d '{
    "concurrency": 3,
    "items": [
      { "instruction": "Hook A: price drop, gift mug" },
      { "instruction": "Hook B: arrives before Dec 24, gift mug" },
      { "instruction": "Hook C: three colours, gift mug" }
    ]
  }'

How do I read the results?

Poll GET /v1/format-run-queues/{id} with a key that has formats:read. The queue status is queued, running or completed, and completed means every item is terminal, not that every item succeeded, so branch on counts.failed. Read each child at GET /v1/format-runs/{run_id}. To stop one, cancel the child with POST /v1/format-runs/{run_id}/cancel; there is no public queue cancel.

How many variants is a sensible first batch?

A bulk queue holds 1 to 100 items and the window is 1 to 16 concurrent runs, so the size of the test is your choice and not a platform limit. A first batch of three to five hooks on one hero SKU is easy to review by eye. Put the same product attachment in every item and change only the hook, so a difference in results can be traced to the hook and not to a different still.

Each run is billed as its own Format run, so set generation_spend_cap_usd on every item. A queue of 20 items with a cap of $15 each cannot spend more than $300 on generation, which is the number to compare against your test budget before you submit.

What are the limits?

• Sume makes the clips; it does not split traffic or report which hook won. Run the split in the ad platform.
• Check that sume-video-hook takes your kind of input with GET /v1/formats/sume/sume-video-hook before you queue a batch.
• Service-account keys are refused for Format runs and bulk queues.
• Google's post gives a rollout window, not a date, so check whether Asset Studio shows the feature in your account.