We measured the cut rhythm of Shorts that work. Here are the numbers.

Most advice about Shorts is written from the outside. Somebody watches a few videos that did well, notices they feel fast, and writes "cut every two seconds" as though that were a measurement.

We had a reason to be more precise. Before our engine could edit a vertical documentary, it needed to know what one is, and "fast" is not a specification. So we took ten Shorts that hold attention, ran every one through frame-accurate analysis, and wrote down what came back. Seven were English, three were Turkish. The numbers below are those files, not an impression of them.

How long is a Short that holds?

Between 36.8 and 58.3 seconds.

Not 15. Not "under a minute, the shorter the better". The floor matters more than the ceiling here: nothing in the reference set was shorter than 36 seconds, and the pieces at the top of the range were not padded, they were carrying more story.

That range became a hard band in our own QA. A render that comes back at 28 seconds fails, and so does one at 70. Both are real failure modes: the short one has skipped the middle of its own story, and the long one has stopped being a Short and become a badly cropped video.

How often do they cut?

Between 8.2 and 25.1 cuts per minute.

Look at how wide that is. The bottom of the range is a cut roughly every seven seconds; the top is one every two and a half. Both of these were in the set, and both worked, which kills the single most repeated piece of Shorts advice in one line. There is no correct cutting speed. There is a correct cutting speed for the thing you are saying.

The pattern underneath is easier to state: the cut rate follows the density of the narration, not the clock. A sentence that names three places in eight seconds wants three shots. A sentence that lands one hard fact wants one shot that stays there while the fact registers.

How long can one shot sit?

Median shot length across the set ran from 2.03 to 6.00 seconds.

The six-second figure surprised us, because the received wisdom says a static vertical shot dies after about two. It does not. It dies when nothing is developing inside it. A six-second shot where a crowd is moving toward the camera is not the same object as a six-second shot of a still photograph, even though a stopwatch cannot tell them apart.

Here is one real render measured against that band. Job b2e4b3eb, a 53.47-second piece about Nadia Comaneci at the 1976 Olympics: 14 cuts, 15.7 cuts per minute, median shot 3.2 seconds, sourced from five separate archives. Every number sits inside the reference range, which is what "within reference" means when our QA stamps it.

What else the reference set agreed on

Three things were true in all ten, with no exceptions, and they turned out to be the ones that are easiest to get wrong in an automated pipeline:

No black frames. Zero. Not a black frame between scenes, not a fade through black, nothing. A single dark frame in a vertical feed reads as the video ending, and the thumb is already moving.

No silence. Also zero. Every gap in the narration is carrying something, even if that something is room tone.

Loudness between -11.5 and -14.8 dBFS. Consistent enough across ten measured Shorts to make it a useful reference band rather than a guess.

Our QA bands were written to contain these measurements rather than to sound reasonable, and there is a rule in the codebase about it: a change that narrows a band tighter than its own evidence fails the build. The measurement outranks the opinion, including ours.

What this means if you are editing by hand

Three things worth stealing:

Stop editing to a metronome. Cut when the sentence changes, and let the rate fall out of that. If you find yourself cutting because it has been two seconds, you are decorating.

Watch your black frames. Most editing software will happily put a two-frame dip between clips and you will never see it in the timeline. It shows up in the retention graph.

Measure your own set. Ten videos took us an afternoon to analyse. If your niche is different from documentary, your bands will be different from ours, and you will learn more from your own ten than from anyone's blog post, this one included.

What this means if you are not editing by hand

This is the specification our engine works to. When ViewMade renders a Short, the output is checked against those bands before it is handed over, and a render that falls outside them is a failed render, not a delivered one. Failed renders return their credits automatically.

The reason we can be this specific about the format is the same reason we can be specific about the footage: everything in the video was filmed by somebody, sourced with attribution, and measured after the fact. The dated production snapshot contains 23 completed renders with a median delivery time of 697.5 seconds, about eleven and a half minutes from topic to finished file.

There is no version of this where you get the numbers by guessing.