32-second event recap: 2 jobs on Wan 3.0 or Seedance 2.5

A 32-second event recap is more than one job on Sume: Wan 3.0 and Seedance 2.5 each stop at 30 seconds per request, so it takes 2 jobs of 16, 16 seconds. At 720p and 9:16 the set costs $4.00 on Wan 3.0 and $18.4896 on Seedance 2.5.

The limits come from Sume's video generation docs (read 2026-10-06): seedance-2.5 accepts 4 to 30 seconds at 480p, 720p and 1080p, and wan-3.0 accepts 2 to 30 seconds. A event recap up to 30 seconds therefore needs no stitching.

What does a 32-second event recap cost by model and resolution?

Both prices below are computed from Sume's rate card on 2026-10-06 for a 9:16 request. Wan 3.0 is a per-second rate by resolution; Seedance 2.5 is priced by video tokens, so its price scales with the pixel count too.

How do I send it?

Submit with POST /v1/videos, send an Idempotency-Key so a retry returns the original job, and poll the polling_url or pass an HTTPS callback_url. Send 2 requests, one per segment, and join the clips afterwards with a Timeline render.

curl -X POST "https://api.sume.com/v1/videos" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: clip-wan-3.0-16s-001" \
  -d '{
    "model": "wan-3.0",
    "prompt": "A product on a desk, slow push-in, natural light",
    "duration": 16,
    "resolution": "720p",
    "aspect_ratio": "9:16"
  }'

What should I check first?

Prices change when providers change theirs, so the table is a snapshot, not a quote.

• Read supported_durations and supported_resolutions from GET /v1/videos/models before you pick a length.
• Sume reserves the provider list price times 1.25 at submit; usage.cost on the finished job is the billable amount.
• Segments are separate generations, so keep the same prompt scaffolding and reference images in each to keep the event recap consistent.