Break down a UGC ad by scene: $0.30 analysis or free stills?

To break down a UGC-style ad you own, start with POST /v1/video-inspect: probe and stills are unbilled and a transcript is $0.01 per audio minute. The legacy video analysis adds typed scenes with shot type and on-screen text for $0.30, but its docs say not to start new work on it.

What does each route give me?

Video inspect returns probe, up to 24 stills (8 mid-bin stills by default) and an optional transcript with words[] and sentence segments[]. The video analyses resource adds a typed scenes[] array. Neither one predicts performance or generates video.

How do I get a hook-by-hook look?

Ask inspect for a still per second with frames: { "fps": 1 }, and for segmentation.mode: "sentence" so the transcript arrives as caption-line-shaped segments. Then read the first three seconds of stills against the first sentence to see what the hook shows and says.

curl -X POST https://api.sume.com/v1/video-inspect \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ugc-breakdown-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/ugc-ad.mp4",
    "frames": { "fps": 1 },
    "transcribe": true,
    "segmentation": { "mode": "sentence" }
  }'

What then?

Use what you learn to write your own script, and render it with a Sume avatar video, which takes a script or scene plan of 4-60 seconds. Only analyze ads you own or have permission to use, and import them to media.sume.com first.