Shorts has no dislike button: test three openings before you post

YouTube announced on June 25, 2026 that it is retiring the dislike button on Shorts and replacing the thumbs-up with a heart. So the cheapest negative signal, a viewer tapping dislike, is gone from the Shorts surface, and you should judge an opening before the audience does. With Sume you can cut three different openings from one 30 second clip for 2 cents each, look at the stills, and post only one.

What the change is, and what it is not

The YouTube Blog post (read 2026-10-05) says YouTube will rely on the controls "Not Interested" and "Don't recommend this channel" to tune a viewer's feed. It does not describe a new creator-facing score, so this post does not claim one. What it means for a workflow is simple: if you used to publish and watch the dislike ratio as an early warning, that warning is no longer there for Shorts, and the decision has to move earlier.

Where Sume fits: pre-publish openings, not post-publish ratings

The YouTube Help page for Shorts (read 2026-10-05) says Shorts can be up to 3 minutes long and that you can upload a Short at a maximum resolution of 1080p. A 30 second model clip is well inside both limits, so the experiment below costs you only trim jobs.

A three-opening test for 6 cents

Take one finished clip that is already on Sume media storage (import it first with POST /v1/media-imports). The video trim API cuts a [start, end) range into a new MP4 and leaves the source alone. Each call is a public rate of $0.02 per job, and precision: exact is a frame-accurate re-encode. Cut three 15 second versions that open on different beats: start at 0, 6 and 12 seconds.

Then run video inspect on each result with frames: { at: [0, 1, 2] }. You get stills as durable media.sume.com image artifacts with t, width and height. Put the three first-second stills side by side. If one opening needs a sentence of explanation, it will not survive a swipe.

for START in 0 6 12; do
  curl -s -X POST https://api.sume.com/v1/video-trim \
    -H "Authorization: Bearer $SUME_API_KEY" \
    -H "Content-Type: application/json" \
    -H "Idempotency-Key: open-test-$START" \
    -d '{
      "video_url": "https://media.sume.com/artifacts/artf_demo/clip.mp4",
      "start": '"$START"',
      "duration": 15
    }'
  echo
done

Read the three results

Each response is a job. Poll GET /v1/jobs/:id/status, then read GET /v1/jobs/:id/result; the result has kind: video_trim and a new video_url. Do not send end and duration together, because the API returns video_trim_range_conflict.

What the test costs

The numbers below come from the Sume docs. They are list rates, so confirm the live price in GET /v1/catalog before you budget a large batch.

Scale up and caption last

If you need to compare more than three, the bulk runs API queues up to 100 Format runs with a concurrency window of 1 to 16. That is the right tool when each variant is a full Format run, not a trim.

Caption only the winner. The video captions API burns captions with one of the documented styles, and a standalone job reserves $0.20 for a clip of up to 60 seconds. Burning captions on losers is the expensive way to learn the same lesson.

Make the three cuts different on purpose

The point of three openings is that they differ in the first two seconds, not in the whole video. Cut them from beats that already exist in the source: a result shot, a problem shot, and a question to camera. Because trim is a plain range cut, every variant keeps the same audio, the same captions timing and the same ending, so the only thing you compare is how the video starts.

Set precision deliberately. The default exact re-encodes (libx264, yuv420p) so the cut lands on the frame you asked for. keyframe is a stream copy that can start a GOP early, which is faster but can pull a few extra frames into the front of your opening. For an opening test, use exact, and read actual_start_seconds in the result if you ever switch.

A rubric for the stills

Sort the stills with a short rubric and write the answer down before you look at any numbers from the platform:

• Does the first still show the subject, or a logo or a blank frame?
• Is any text in the first second readable at phone size?
• Does the second still change the picture, or is it a near copy of the first?
• Would you stop scrolling on this frame if you had never seen the video?

Reuse the winner elsewhere

The three-cut test is also useful for a Reels or TikTok upload, since the same clip can go to all three. The Reels three-second hook post shows the same still-based check from the Instagram side.

The takeaway

Without a dislike button you lose a signal, not a decision. Make the decision before posting: three openings, three stills, one winner, and only then a caption job.