Three app previews show in search: queue three hook variants at once

Apple lets you upload up to three app previews to a product page, and up to three of them may appear in search results, so the three previews you choose are your whole search-results pitch. The practical move is to produce three deliberately different hook variants of one app video and pick by what a viewer sees in the first second.

The counts come from App Store Asset Best Practices and Resources (read 2026-10-11). Sume can queue the three variants as one bulk run of its mobile-app Format; it does not choose which one Apple shows or measure which wins.

The counts Apple states

Both rows are quoted from the Apple page.

Three hooks, not three trims

Autoplaying previews repeat and are usually muted, so a search-results viewer decides on the opening seconds. Three trims of one idea waste the slots. Vary what the first shot proves.

• Variant A opens on the problem the app solves.
• Variant B opens on the finished state, the one screen the user wants.
• Variant C opens on a before and after, with no talking.

One bulk run for all three

The Format sume-mobile-app-ugc is a first-party catalog entry whose example input asks for a 9:16 mobile-app UGC ad where a creator holds a phone and shows a budgeting app. Bulk runs says a queue takes 1 to 100 items, concurrency of 1 to 16, one Idempotency-Key per batch, and that each item is the same body as a single run, so each item can carry its own generation_spend_cap_usd.

The request below sends three items with a $20 cap each, at concurrency 3. Read the Format's GET /v1/formats/sume/{slug} first, since Format catalog says that call shows its description and io profile. The data in a 202 holds the queue; poll its status_url, and remember that completed means every item is terminal, not that all succeeded.

import os, uuid, requests

hooks = [
    "Hook A: creator gasps at a hidden subscription total",
    "Hook B: creator shows the one-screen budget first",
    "Hook C: creator compares before and after the app",
]
items = [
    {"instruction": h, "generation_spend_cap_usd": 20}
    for h in hooks
]
r = requests.post(
    "https://api.sume.com/v1/formats/sume/sume-mobile-app-ugc/bulk-runs",
    headers={
        "Authorization": "Bearer " + os.environ["SUME_API_KEY"],
        "Idempotency-Key": str(uuid.uuid4()),
    },
    json={"concurrency": 3, "items": items},
    timeout=60,
)
print(r.status_code, r.json()["data"])

What a batch of three does and does not guarantee

A fresh Idempotency-Key per batch matters: per the docs, replaying a spent key returns the old queue, not a new one. The spend cap bounds each item; it is a ceiling and not a price quote, so read the usage block on the finished runs for what was billed.

Choose the three finalists by watching them muted, at phone size, with the first second only. Then format each to Apple's specs outside Sume; this post did not read those. For looping behavior see App preview video that loops and works muted, and for the ad-versus-preview split see Mobile-app UGC Format: holiday app ads vs App Store previews.