Headline card above an app screen recording: compose stack, $0.36

An app promo with a headline above a screen recording is one POST /v1/timeline-1.0/compose call with operation: "stack", and it costs $0.02 per job. Three headline variants over the same recording cost $0.36 once you add a $0.10 image at 1K for each headline.

The cost per variant

Compose has no provider inference; it runs ffmpeg on the worker, so the rate is flat at $0.02. The headline image is the only variable. Each variant is a new job.

Request

layout.image_region is top and ratio is the still's share of the frame (0.1 to 0.9). With ratio: 0.3 the headline takes the top 30 percent and the recording the rest. The output is 1080x1920 by default; set output to 720x1280 to match a later Timeline render. Both URLs must be media.sume.com artifacts of your workspace, and the still must probe as an image.

curl -X POST https://api.sume.com/v1/timeline-1.0/compose \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: app-promo-headline-a" \
  -d '{
    "operation": "stack",
    "image": { "url": "https://media.sume.com/artifacts/artf_demo/headline-a.png" },
    "video": { "url": "https://media.sume.com/artifacts/artf_demo/screen.mp4" },
    "layout": { "split": "horizontal", "image_region": "top", "ratio": 0.3 },
    "output": { "width": 720, "height": 1280, "fps": 25 }
  }'

Limits and gotchas

• The output length comes from the video layer; the still cannot make it longer. The ceiling is 300 seconds.
• A recording with no audio is only a warning (compose_video_has_no_audio); add a voice or music at the Timeline step.
• Stack and overlay vocabulary cannot be mixed in one request.
• Rendered text inside an AI image can be wrong. Proofread each headline before you publish.