YouTube video A/B test: how to make three hook cuts

To make three cuts for a YouTube video A/B test, start all three from the same finished take and change only the opening: pick three different in-points, keep the same out-point, and export each range as its own file. On Sume, video trim does that for $0.02 per job, so three variants cost $0.06 before any captions.

YouTube's announcement lists video A/B testing, up to three cuts to test hooks, under "coming soon", and this page quotes nothing else about it. The Sume figures come from the docs page linked above; both were read 2026-09-29.

What has YouTube said about testing three cuts?

Only the headline: the Studio section of the Made on YouTube post says that video A/B testing, up to three cuts to test hooks, is coming soon. It also says creators have run more than 40 million title and thumbnail experiments since 2024. The post, as read, gave no rules for the cut test, so none are assumed here. Prepare files that would fit any reasonable rule: same length, same ending, one difference.

What should differ between the three cuts?

Only the first seconds. If the ending, pacing and captions also change, a winner cannot be traced to the hook. Choose three moments in the raw take that could each open the video, for example a question, a result shown first, or a blunt claim, then cut each from its own start to one shared end.

How do I cut them on Sume?

Import the take into your workspace first, because trim reads only media.sume.com files. Then send one POST /v1/video-trim per variant with the same end and a different start. Each call needs its own Idempotency-Key, and the result is a new MP4; the source is untouched. The default precision of exact is a frame-accurate re-encode, so a hook that starts mid-sentence starts where you asked.

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-b" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "start": 6,
    "end": 42
  }'

Can I burn captions on each cut?

Yes, with video captions on each output, but do it for all three or none. Styles named in the docs include slam, punch and tiktok-green. A silent clip fails as caption_no_speech; pass cues to burn authored text instead.

What does this not do?

Sume makes the files. It does not upload them to YouTube, run the test, or read results; that stays in YouTube Studio once the feature ships. For a cut-and-join approach with a shared body, see hook variations for UGC ads.