A/B test three hook lines on one Short with caption cues

If you want to compare three on-screen hook lines on the same Short, keep one clean render and burn each line onto it with Sume video captions using cues. Each variant is one $0.20 caption job, and you do not re-render the video. YouTube has said that testing up to three video cuts is coming soon in Studio.

What YouTube announced

At Made on YouTube 2026, YouTube said video A/B testing is coming soon and will let creators test up to three video cuts to find the best-performing hooks, and that creators have run more than 40 million A/B experiments since the 2024 launch (YouTube Blog). The announcement gives no date and no eligibility, so plan the files now and check Studio before you rely on it.

One clean clip, three lines

POST /v1/video-captions takes a public HTTPS video_url plus cues with text, start and end in seconds. With cues, Sume runs no speech-to-text and burns your text at those times. You can send only one of script_text, words, cues and segments, so a cue job burns the hook line and not speech captions. That fits a Short with music or b-roll; for a talking Short, test the opening in the edit instead.

{
  "video_url": "https://example.com/clean-short.mp4",
  "style": "punch",
  "cues": [
    { "text": "Stop scrolling: this takes 20 seconds", "start": 0.0, "end": 2.5 }
  ]
}

Hold everything else equal

Send the same body three times with a different text and a different Idempotency-Key each time. Hold the style, timing and position constant so the line is the only change. Keep the first frame the same, because the thumbnail may be what viewers see first.

Three variants of one clip:

Keep the test clean

Use short lines that fit the safe width and read at a glance. Change one idea at a time: a question, a claim, a number. After the test, keep the winner and drop the rest. Sume prepares the files; the experiment, the traffic split and the result stay in YouTube Studio.