AI video generator for Instagram Reels: 9:16 models and durations

For a 9:16 Reels clip on Sume, send aspect_ratio: "9:16" to POST /v1/videos and pick the model by how long the clip must be. seedance-2.5 accepts 4 to 30 seconds and wan-3.0 2 to 30; minimax-h3 runs 5 to 15, and gemini-omni-flash-1.1 3 to 10. sume/auto also takes 9:16 at 3 to 10 seconds.

The Video generation docs, read 2026-09-29, say limits are not uniform, so read supported_durations and supported_aspect_ratios from GET /v1/videos/models before you submit. This post does not state Instagram's own Reels length rules; check those on Meta's help pages.

Which models take 9:16, and for how long?

How do I make a vertical clip of a given length?

Set model, aspect_ratio and duration in one request, and send an Idempotency-Key so a retry returns the original job. Video generation is asynchronous: you get a job id and a polling URL, then poll until the status is completed.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: reels-vertical-001" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "A vertical product clip on a desk, natural light",
    "aspect_ratio": "9:16",
    "duration": 15
  }'

When should I pick auto instead of a pinned model?

Use sume/auto when a short clip is enough and you do not care which family runs: it defaults to 720p and 8 seconds and is capped at 10 seconds. Pin a model when you need more than 10 seconds, since auto has no longer setting. See the auto duration limit for the details.

What if a model does not list 9:16?

Not every catalog model advertises an aspect ratio. Generate at a ratio it does list and reframe afterwards; reframe after generation covers that. Also note size returns 400 unsupported_parameter on every v1 model, so use resolution plus aspect_ratio rather than pixel dimensions.

How do resolution and audio limit the choice?

Resolution changes by model. seedance-2.5 offers 480p, 720p and 1080p, and gemini-omni-flash-1.1 offers 360p, 720p, 1080p and 4K with native synced audio. minimax-h3 runs at native 480p and 768p, where 768p is first-class rather than 720p. Check supported_resolutions in the models response for the id you pin.

If the clip needs an audio track, the models list reports generate_audio per model, so read that field rather than assuming. Billing is reserved on submit at provider list times 1.25 on every model, so a longer 30-second clip costs more than a 15-second one on the same model.