No dislike on Shorts: how to test three hooks with pinned Sume ids

With the dislike button gone from Shorts, you can no longer read a thumbs-down as a verdict on your hook, so the practical test is to publish variants and compare results. Make three versions of the first few seconds on one pinned Sume video id, change one thing between them, and keep everything else identical.

The change itself is on YouTube's own page. The YouTube Blog post on the Shorts experience (read 2026-10-10) is dated June 25, 2026. It says YouTube is replacing the thumbs-up with a heart, "retiring the dislike button" in favor of more granular controls, and giving viewers "Not Interested" and "Don't recommend this channel" to shape their feed. The same post adds a clear screen mode that hides icons and text, a way to double playback speed, and a Shorts timer that can be set to zero.

What the page does and does not tell creators

The post is written for viewers. It says what viewers can tap, and I found nothing in it that says creators will see a count of Not interested taps or of Don't recommend this channel taps. So do not plan around a replacement for the dislike number. Plan around the numbers you already have for each upload, and make sure the uploads are comparable.

Why pin the video id for a hook test

Sume's Video Generation page says model: "sume/auto" lets Sume select the family, and that Sume never discloses which family served the request. That is convenient for a one-off clip and wrong for a test, because three variants could come back from three different families and look different for reasons unrelated to your hook. Name the id yourself in all three requests.

• Pin gemini-omni-flash-1.1 if you want 3 to 10 seconds in 16:9 or 9:16 with native synced audio, per the Video Router page.
• Send the same aspect_ratio (9:16), resolution and duration in every request.
• Give each variant its own Idempotency-Key, such as hook-a, hook-b, hook-c; the docs say a replay with the same key returns the original job, so reusing one key would give you the same clip three times.

Change one thing at a time

A hook test is only readable if the variants differ in one respect. Good single changes are the first action (a close-up of the product versus a person reacting), the opening sound, or the first line of on-screen text. Bad changes are anything that alters several at once, such as swapping model, framing and pace together.

Because the first seconds are what you are testing, you can generate only the opening: three 4-second or 5-second clips, then attach the same body to each. Sume's Timeline 1.0 page describes assembling ordered video slots with one audio spine into a single MP4, so the shared body can be identical across all three uploads.

Remember that viewers who double playback speed hear your opening twice as fast, and clear screen mode removes interface icons and text from view. Neither changes how you generate, but both are reasons to check the hook with sound on, with captions burned into the picture, and at normal and doubled speed before you publish.

Reading the result honestly

Publish the three uploads under similar conditions and compare the numbers YouTube already gives you for each, such as views and watch behavior. Small differences on one run are noise. If one hook wins by a wide margin, repeat with a new body before you trust it. Sume's contribution here is only that every variant was produced with the same request shape, so a difference in results is not a difference in tooling.