How do I turn an app screen recording into a Black Friday sale ad?

To turn an app screen recording into a Black Friday ad, trim the best 15 seconds with POST /v1/video-trim, render it over a music bed with Timeline 1.0, and burn the sale offer into it with caption cues. On Sume that is $0.02 for the trim, $0.125 for the music, $0.10 for the render and $0.20 for the captions: $0.445. None of it uses a generative video model, so the product on screen is exactly your product.

That matters for software. A generated clip would invent an interface. A recording shows the real screen, and the only new things in the ad are the cut, the sound and the words.

Pick the 15 seconds

Watch the recording once and write down the second where the payoff starts: the report appearing, the export finishing, the checkout completing. Start two seconds before it. The trim call takes video_url, start and exactly one of end or duration, and it returns a new file; the source is never changed. Import the recording first with a media import, because every URL must be a media.sume.com file in your workspace.

A trim is a flat price, so there is no saving in trimming shorter. Choose the length for the placement. Vertical feeds reward 15 seconds or less, and the ad needs three beats: what it does, what the offer is, where to go.

curl -X POST https://api.sume.com/v1/video-trim \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: bf-app-trim-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/screen.mp4",
    "start": 41.5,
    "duration": 15
  }'

Add the bed and the sale text

Generate the bed with Music Router. For software, ask for something steady and bright with a clear lift at the offer: 'A 15-second upbeat electronic track, 118 BPM, E minor to G major at 0:10. Plucked synth, soft kick, light claps. Instrumental, no vocals.' It is a flat $0.125 whatever the length. Then render with the trimmed clip in a single video[] slot, the music as audio.url and duration_seconds: 15; the default output is 1080 by 1920.

Burn the offer with a caption job using cues: 'Black Friday: 40% off annual plans' from 0 to 15 seconds is too long to read as one line, so split it: the benefit from 0 to 5, the discount from 5 to 11 and the end date from 11 to 15. A standalone caption job is $0.20 for up to 60 seconds. Cues are plain strings you control, so there is no spelling risk in the numbers.

Why a recording beats a generated clip for software

A generative model does not know your interface. It will draw buttons that look plausible and labels that say nothing, and a buyer who knows the product will see it at once. A trimmed recording is the product. The ad then spends its budget on the part a recording cannot give you: a sound bed that sets the pace, and clear words about the offer.

It also makes variants cheap. One recording gives you a feature cut, a price cut and a deadline cut. Each one differs only in its start time, its caption cues and perhaps its music, and none of them needs a generation queue. If the recording has private data on screen, blur or crop it before you import it, because everything you import is hosted on a Sume media URL.

Keep the recording at the size you will publish. The render's default output is 1080 by 1920, so a landscape desktop capture is placed with fit: cover, which crops the sides. Use fit: contain or blur if the whole screen must stay visible, and check that the text on screen is still readable at phone size.

What it costs and the variants

• Three offers, same recording: reuse the trim and music, then pay $0.30 per variant for a render and a caption pass: $1.045 for three.
• A second cut for Cyber Monday (November 30): another trim at $0.02, and re-use the same bed.
• Keep the Idempotency-Key tied to the variant name so a retried request never bills twice.