Instagram's three Reel length tiers: cut one talk into 15, 60, 180 s

Instagram's own Reels guide says a Reel can run from 3 seconds to 3 minutes and splits that range into three tiers: 3-15 seconds, 15-60 seconds and 60 seconds to 3 minutes. If you have one talk or demo, you can cut it into one clip per tier with three Sume video trim calls, then post the version that fits the job.

The tiers come from Instagram's How to Make Instagram Reels: Complete 2026 Guide, read on 2026-10-03. The guide is a creator tip sheet, not an API or ad spec: it does not state an aspect ratio, a safe zone or which tier reaches more people, so this post does not claim any of those.

What does Instagram say each length is for?

The guide pairs each tier with a content type. That pairing is the only thing it says about length, so treat it as a planning aid rather than a ranking rule.

How do I cut one source into three Reels?

Video trim takes one clip that is already on the Sume media host, a start and exactly one of end or duration, and returns a new MP4. The source is untouched, so you can run all three cuts against the same file. Import the source first with POST /v1/media-imports; the trim route does not fetch from the open internet.

Every write needs an Idempotency-Key. Use a different key per cut, or the retry protection will treat them as one request. The default mode is async, so the call returns a job and you poll GET /v1/jobs/:id/status and then GET /v1/jobs/:id/result; see Jobs and results.

curl -X POST https://api.sume.com/v1/video-trim \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: talk-tier-15s" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "start": 42,
    "duration": 15
  }'

How do I choose where each cut starts?

Sume does not pick the best moment for you. Video inspect can return eight sampled stills, a probe and, with transcribe: true, a transcript. Add segmentation.mode: "sentence" and you also get gapless sentence segments, which are caption-line shaped and give you clean in and out points that do not cut a word in half.

Read the sentence list, then pick three starts: a single strong sentence for the short tier, a full argument for the middle tier and a continuous section for the long tier. Transcription is billed at $0.01 per audio minute on the docs page, and the docs tell you to confirm live prices in GET /v1/catalog. A silent clip fails with inspect_source_has_no_audio, so check probe.has_audio first.

What limits does the trim step have?

Video trim accepts a source up to 1800 seconds and returns an output of 0.2 to 900 seconds, so all three tiers sit well inside the cap. The default precision is exact, a frame-accurate re-encode. The keyframe option copies the stream and can start a little early, which is why the result reports actual_start_seconds. For Reels that need to start on a specific word, keep exact.

Each trim job is priced at $0.02 on the docs page, so a three-tier set of cuts costs six cents in trim fees, plus any transcription. If you ask for an end past the source, it clamps and the result warns trim_clamped_to_source.

What does Sume not do here?

Sume does not publish to Instagram from this flow, does not decide which tier will perform, and does not read your Instagram analytics. After trimming, you still need to add captions, audio and a cover in the app or through your own publishing step. A practical way to use the three versions is to post the 15-second cut first, check retention in your own Insights, and only then decide whether the 60-second or 3-minute version is worth posting.

Keep a note of which start and duration produced each file. Because trim never changes the source, you can re-cut a different window later without re-importing anything.