Seedance 2.5 multi-shot prompt: timestamps, shots and one example

To get several shots in one Seedance 2.5 clip, split the prompt by time ranges, one shot per range, and describe each with subject, action, scene, style, camera and sound. ByteDance's launch post says the model handles multiple logically connected shots; BytePlus's prompt guide shows a shot-by-shot prompt with time ranges running to 30 seconds.

Send it to seedance-2.5 in POST /v1/videos with duration matching the last timestamp.

What should each shot include?

The tutorial's formula is subject + action + scene + style + camera + sound. Its prompt guide advises keeping multi-panel storyboard images to 15 panels or fewer.

What does the request look like?

Keep each range to one idea, and keep the subject description identical across shots so the character stays consistent.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "0-4s: wide shot of a lighthouse at dawn, slow push-in, waves. 4-8s: a keeper in a yellow coat climbs the stairs, tracking shot, boots on iron. 8-12s: close-up of the lamp igniting, warm glow, low horn.",
    "duration": 12,
    "resolution": "720p",
    "aspect_ratio": "16:9"
  }'

How many shots fit in one clip?

Neither page states a shot cap. A workable rule is 3 to 4 seconds per shot, so 30 seconds holds about eight. If shots need different references, generate them separately and join the clips; see multi-shot video generation.

What if one shot misses?

Regenerate only that shot as its own short clip rather than the whole 30 seconds; see retry one scene of an AI video.