Shorts views count every start and replay: three cut lengths for $0.06

YouTube's Shorts help page (read 2026-10-08) says that starting March 31, 2025, a Short view is counted as the number of times it starts to play or replay, with no minimum watch time. That makes the view count a poor way to compare two cuts of the same clip, because a 2 second look counts the same as a full watch. You can still make several cut lengths cheaply with Sume video trim: three variants at $0.02 each cost $0.06 in total.

What the page says and does not say

The page is short on mechanics. It gives the view definition, the 3 minute limit and the 1080p limit. It does not describe retention measures, so any comparison beyond views has to come from your own analytics screens, not from this page.

Three cuts from one source

Send three trim jobs on the same media.sume.com source, each with start 0 and a different duration. Each job needs its own Idempotency-Key. Exact precision gives frame-accurate lengths. The cost is 3 x $0.02 = $0.06, and the check is the duration_seconds field in each result.

How to compare them

Publish the variants as separate Shorts and compare them on measures that are not the raw view count, using what YouTube's analytics shows you. Keep everything else the same: title style, posting hour, and thumbnail frame. The trim keeps the geometry and audio of the source by default, so the only change between A, B and C is length.

If a variant needs the same length but a different first second, trim with a different start and the same duration. A trim with start 2 and duration 10 is as cheap as one with start 0. The longer-source limits are in the 180 second trim post. For a cost view of longer series, see this post.

Keeping the test fair

A cut-length test only means something if the cuts differ in length and nothing else. Use the same source, the same start, and exact precision, so the frames match. Name each Idempotency-Key after its variant, for example short-5s, short-10s and short-15s, so a retry cannot silently change which variant a key maps to.

Exact precision re-encodes with libx264 in yuv420p, while keyframe precision copies the stream and can begin a GOP early. For a fair comparison choose exact, which is the default. Each result reports duration_seconds, and it is worth checking that A, B and C are close to 5, 10 and 15 before you publish.