Amazon Ads Video Generator or a Format API for product ads?

Use Amazon Ads' Video Generator when the ad lives on Amazon and you build it by hand in Amazon's console. Use a Sume Format run when the product video has to come out of your own pipeline, for any channel: one POST /v1/formats/sume/{slug}/runs call starts a finished video, and you collect it by webhook or by polling. The two overlap on the first step (a product in, a video out) and differ in where the work happens and who can trigger it.

What does the Amazon Ads Video Generator take in and return?

Amazon's announcement page describes a Video Generator that makes realistic product videos showing items in everyday use, with multi-scene output, text animations and background music. It returns six video options per request, and it can turn an image into a short GIF-style motion clip. Inputs are product images, product page data or an existing ad.

The same page lists an AI Creative Studio (beta, U.S. advertisers only) that combines image and video generation, and an Audio Generator (beta, Amazon DSP advertisers only) that writes 30-second audio ads. Those limits matter: the tooling is scoped to Amazon advertisers.

What does a Sume Format run take in and return?

Sume ships a catalog of Formats that answer at the reserved sume handle. For product ads, three fit: sume-product-commercial (a premium hero product film), sume-product-usage-demo (one real action, texture or result in a casual setting) and sume-close-camera-ugc (a handheld selfie-style recommendation). Any key with the formats:write scope can call them, and the run and its spend belong to the key that made the call.

Before you call one, read it with GET /v1/formats/sume/{slug}. The io profile tells you the input kind (url, text, image or product) and the output kind. The run itself is the call below, with a per-run spend cap so a bad input cannot run away.

curl -sS -X POST "https://api.sume.com/v1/formats/sume/sume-product-usage-demo/runs" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: sku-1042-demo-v1" \
  -d '{
    "instruction": "Vertical 9:16, hands-on demo, no music.",
    "input": { "product_url": "https://shop.example.com/p/1042" },
    "generation_spend_cap_usd": 10,
    "communication": { "webhook_url": "https://example.com/hooks/sume" }
  }'

How do the two compare?

How do you pick?

If every ad runs on Amazon and your team works in Amazon's console, the built-in generator saves a hop. If the same product needs a video for your own store, a TikTok ad and a marketplace listing, or you want a SKU sheet to turn into videos overnight, a callable Format is the better fit, because the trigger is code and the receipt is machine-readable.

A run takes minutes, not seconds. It moves through queued, processing and then completed, failed, canceled or skipped. Send a webhook URL, or poll GET /v1/format-runs/{run_id} with a back-off, and stop on a terminal status.

• Pick the Format by the shot you need: hero film, hands-on demo or creator-style recommendation.
• Set the spend cap before you scale past a handful of SKUs.
• Reuse one Idempotency-Key per SKU and version so a retry never pays twice.