Instagram trial reels: test three hooks from one source clip

To test hooks with trial reels, cut three versions of the same footage that differ only in their first seconds, publish each as a trial reel, and graduate the one that wins. Meta defines trial reels as reels shared only to non-followers, and the API takes a trial_params.graduation_strategy of MANUAL or SS_PERFORMANCE. Sume's video trim makes the three cuts for $0.02 each.

Facts about trial reels come from Meta's Content Publishing guide and IG User Media reference, both read 2026-10-02. Meta does not publish a performance threshold for SS_PERFORMANCE on those pages, so this post does not invent one.

What does a trial reel API call need?

A trial reel is a normal Reels container plus trial_params. Meta says media_type must be REELS if trial_params is sent, and graduation_strategy is required inside it. MANUAL lets you graduate in the native app. SS_PERFORMANCE graduates automatically if the trial reel performs well.

For a hook test use MANUAL. Automatic graduation would publish more than one variant of the same footage to your followers, which defeats the point of the test.

How do I cut three hooks from one clip?

Video trim takes one Sume-hosted clip, a start, and exactly one of end or duration, and returns a new MP4 holding only that range. The default precision is exact, a frame-accurate re-encode with libx264 and yuv420p. Output duration runs from 0.2 to 900 seconds, and each request needs an Idempotency-Key.

Give each variant its own key so a retry never queues a second cut. Choose three start points that open on different moments: the result, the problem, or the surprise. Keep the length identical so the only variable is the hook.

for n in 1 2 3; do
  case $n in 1) S=0;; 2) S=14;; 3) S=31;; esac
  curl -X POST https://api.sume.com/v1/video-trim \
    -H "Authorization: Bearer $SUME_API_KEY" \
    -H "Content-Type: application/json" \
    -H "Idempotency-Key: hook-test-$n" \
    -d "{\"video_url\": \"https://media.sume.com/artifacts/artf_demo/talk.mp4\", \"start\": $S, \"duration\": 20}"
done

What should stay the same across variants?

A test only tells you something if one thing changes. Hold the caption, cover, audio, length and posting window constant, and vary the opening alone.

Meta's Reel minimum is 3 seconds, so a short hook still leaves room. If your variant cuts below that, the container will not be accepted; see the 3-second minimum post.

• Same duration on all three trims.
• Same caption, cover_url or thumb_offset, and audio_name.
• One trial_params strategy for all three, set to MANUAL.
• Record the job id and variant name together before publishing.

How do I read the result and pick a winner?

Poll each trim through the job envelope until it is result_ready; the result carries a new video_url, duration_seconds and actual_start_seconds, which is how you confirm the cut began where you asked. See jobs and results.

Then decide the winner in the Instagram app, using the numbers Instagram shows for the trial reel. Sume does not read trial reel performance and does not graduate reels. To audit why a hook worked, video inspect can return stills and a transcript of each variant, which helps you see what the first three seconds actually showed.

What can go wrong with a hook test?

The usual trap is testing too many things at once. A different cover, caption or sound will confound the result. The second trap is a source clip that is too short: three 20-second cuts need at least 31 seconds plus 20 seconds of material after the latest start, so check the source length first with a probe.

The third is reading too much into a small sample. A trial reel reaches non-followers only, so numbers can be small and noisy. Treat the result as a signal for what to try next, and keep the winning cut and its job id so you can reproduce it.