Seedance 2.5 camera move test grid: six 4-second probes at 480p

Before you spend about $17 on a 30-second Seedance 2.5 clip at 720p, run six 4-second probes at 480p, one per camera move: dolly-in, orbit, crane up, handheld, whip-pan and static. Each costs about $1.07 on Sume, so the grid is about $6.45 in total. Keep subject and lighting identical so the only variable is the camera instruction.

Why 4 seconds and 480p

Four seconds is the shortest duration seedance-2.5 accepts on Sume, and 480p is its cheapest resolution, per the Video Router docs. BytePlus lists 480p, 720p and 1080p at 24 FPS (BytePlus: Dreamina Seedance 2.5), so a draft shares its frame rate with a final. There is no seed field to repeat a take, so a probe tells you what the phrase tends to do, not what a later run will do exactly.

The six prompts

Use one sentence for the scene and one for the camera. Do not change anything else between probes.

• Dolly-in: the camera moves slowly toward the subject.
• Orbit: the camera circles the subject at a steady speed.
• Crane up: the camera rises and reveals the whole set.
• Handheld: slight natural camera shake, following the subject.
• Whip-pan: a fast pan to a second object, then settle.
• Static: locked-off camera, only the subject moves.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: cam-grid-orbit-001" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "A ceramic mug on a wooden table by a window. The camera circles the mug at a steady speed.",
    "resolution": "480p",
    "duration": 4,
    "aspect_ratio": "16:9",
    "mode": "async"
  }'

What the grid costs

Use a distinct Idempotency-Key for each probe; reusing a key for a different payload returns 409 idempotency_conflict per the generation admission docs. On a Free plan only one job processes at a time and five more can queue, so six probes fit exactly.

Score the results

Score each probe on three yes or no questions: did the camera do what the phrase says, did the subject stay the same, did the background stay stable. Pull the first and last stills with video frames (the clip must already be stored on media.sume.com, so import a downloaded MP4 first) to compare. Write the winning phrase into your 30-second prompt for the matching beat, as in the beat sheet approach. The results are one sample each, not a benchmark.