How often should you cut in a YouTube video? Pacing measured on real channels
Advice about pacing usually comes as a single number: cut every few seconds, never hold a shot too long. I measured three established narrated documentary channels frame by frame, and they disagree with each other by a factor of about six. Here are the numbers, how they were measured, and what to take from them for your own channel.
How often should you cut in a YouTube video?
There is no single right cut rate. On three narrated documentary channels I measured, the picture changed anywhere from about 2 times a minute (Librarian, 29.4-second median shot) to about 13–15 times a minute (Philip Thompson and The Interpreter, 4.1–5.0-second median shots). Match the format and the reference your audience already watches.
The rest of this page is the evidence. All three channels tell long, narrated, true stories, and two cover the same subject, spies. If there were one correct pace for documentary YouTube, these three would sit close together. They don't.
Which channels were measured, and why these three?
I picked them because they sit at opposite ends of the same genre, which makes them useful as a map of what is possible:
- Philip Thompson, “The Unbreakable Spy of WW2”. A reference channel analyzed from its public video, not a client.
- Librarian, “The Real James Bond: The True Spy You’ve Never Heard Of”. Also a reference channel analyzed from its public video, not a client.
- The Interpreter, a faceless channel that tells aviation, maritime and rail disaster stories with stickman characters. This one is a client: I edited a long-form Tenerife 1977 deep dive for them. The measurements come from two of the channel's own public uploads, a Turkish Airlines disaster compilation and a dam-disaster compilation, first 600 seconds of each.
What do the three channels look like side by side?
| Channel | Picture changes per minute | Median shot | How shots join | Text on screen | Integrated loudness |
|---|---|---|---|---|---|
| Philip Thompson | 12.8 detected, about 14.7 corrected (10.4 full-frame) | 4.1 s | ~33% cuts, ~28% dissolves, ~36% dips to black, 2.6% flashes | 19.6% of the time | −13.7 LUFS |
| Librarian (first 30:34) | 2.16 | 29.4 s | 95.5% cross-dissolves of ~2 s | None added | −15.8 LUFS |
| The Interpreter (client) | 12.9–13.6 | 4.3–5.0 s | Hard cuts (26 of 26 checked) | 34–42% of the time | −22.5 LUFS |
The two spy channels are the sharpest contrast: same subject, and one changes the picture roughly six times as often as the other. The difference is not quality but what each puts on screen: one cuts in moving archival footage and quick portraits, the other lets a single still photograph breathe for half a minute.
How does Philip Thompson pace a spy documentary?
| Shot length | Shortest 0.6 s · 25th percentile 2.6 s · median 4.1 s · 75th percentile 6.2 s · longest 22.6 s |
| Changes per minute | 12.8 detected automatically, about 14.7 after correction; 10.4 counting only full-frame changes |
| Transition mix | Hard cut ~33% · dip to black ~36% · cross-dissolve ~28% · white flash 2.6%. Dissolves last about 0.75 s, dips about 0.8 s |
| Text on screen | 2.0 appearances a minute, 19.6% of the time, 6 s median on screen (names, places, dates, quotes) |
| Movement | Zoom-in 68% of the time; real moving footage 20.6% of the time, archival footage alone 12.5% |
| Audio | −13.7 LUFS integrated, loudness range 3.0, true peak −1.2 dBFS. Music only in the cold open (about 0:10–1:30) and the close; no sound effects |
The most useful number here is not the average. It is the curve. The pace changes across the video:
- Cold open (first minute): about 19 changes a minute. The 61-second open starts in the middle of the story.
- Minutes 0–10: 14.5 a minute.
- Minutes 10–25: 12.5 a minute.
- Minutes 25–30: 10.8 a minute.
- Closing: 8.2 a minute.
The edit spends its fastest cutting where viewers decide whether to stay, then relaxes as the story earns their attention. Portraits change roughly every 5 seconds in stretches with faces, so even the slower middle section rarely sits still. If you read your retention graph and see an early drop, the opening pace is one of the first things to look at.
How does Librarian hold attention with 30-second shots?
| Shot length | Median 29.4 s (66 shots in 0:00–30:34) |
| Changes per minute | 2.16 |
| Transition mix | Cross-dissolve of about 2.0 s on 63 of 66 changes (95.5%), 2 fades to black in the first minute, 1 hard cut in the whole stretch |
| Text on screen | None added: no titles, labels or dates. The only text is inside the images themselves (posters, a magazine) |
| Movement | Slow zoom-out 69% of the time, median 0.68% a second, about 16% of travel per shot. No real footage: every image is a still |
| Audio | −15.8 LUFS integrated. Continuous voice at about 157 words a minute; music bed about 31 dB under the program. No sound effects |
Two changes a minute would feel slow in most formats. Here it works because the narration does the carrying. The voice runs without breaks at about 157 words a minute, the image keeps drifting outward, and each dissolve takes about two seconds, so the picture never actually stops.
One more detail worth knowing: from 30:34 to 45:37, a third of the video, the screen is a single 18-second AI-animated clip repeated about 50 times while the voice continues. That is why the table above only covers the first 30:34.
How does a faceless disaster channel pace its videos?
| Shot length | Shortest 0.4–0.9 s · 25th percentile 2.4 s · median 4.3–5.0 s · 75th percentile 6.0–6.6 s · longest 11.4–13.5 s. Between 2 and 7 shots over 10 s in every 10 minutes |
| Changes per minute | 12.9–13.6 |
| Transition mix | Hard cuts: 26 of 26 checked frame by frame changed in a single frame. No fades found |
| Text on screen | 8.2–9.0 text callouts a minute, 2.5 s median each; text on screen 34–42% of the time, not counting the fixed title banner |
| Material | Archival photos 66% · video b-roll 24% · drawn cards 3.6% · maps and diagrams 3% · date cards 1.8%. Zoom-in 77%, no movement 18.1% |
| Audio | −22.5 LUFS integrated, true peak −5.5 dBFS, loudness range 2.5–2.8. Constant music bed about 24 dB under the narration, no ducking. Longest gap in the voice 1.5 s |
The cut rate lands close to Philip Thompson's, but the way it feels is completely different. Every change is a hard cut, and the energy comes from short white text callouts that pop in and out on top of the picture, 8 to 9 of them a minute. Nothing is ever silent: the music bed never drops out and the voice never pauses for more than a second and a half. The opening packs 10 to 12 shots into the first 30 seconds.
This is what pacing looks like on a faceless channel: with no presenter on screen, the picture and the text have to carry every line, so changes come often and stay short.
What should a creator take from these numbers?
- Pacing depends on format, not on a rule. Two changes a minute and fourteen changes a minute both work inside the same genre. Copying a cut rate from a different format is how an edit ends up feeling wrong without anyone knowing why.
- Measure a shot-length distribution, not just an average. The Interpreter's median is under 5 seconds, yet it still holds 2 to 7 shots past 10 seconds in every 10 minutes. Variety inside the pace is part of the style.
- The transition is part of the pace. A two-second dissolve every half minute reads as calm. Hard cuts at the same rate as a dissolve-heavy edit feel faster. When you describe a reference to an editor, say how shots join, not only how often.
- Pace is a curve. The fastest cutting in the Philip Thompson video is the first minute, and it slows by more than half by the end. A flat rate from open to close is a choice, not a default.
- Text and sound count as stimulus too. The Interpreter's callouts and constant music bed do some of the work cuts do elsewhere. Librarian uses no added text at all and leans on a continuous voice.
Do these numbers apply to gaming or vlogs?
No, and I'd be careful with anyone who hands you one number for every format. This data covers narrated documentary channels only. Gaming, commentary and vlogs have their own pacing, and I haven't published measurements for them here. What does transfer is the method: pick two or three videos your audience already watches in your niche, count how often the picture changes, note how shots join and how much text is on screen, and write that down before anyone edits. That turns "make it feel like this channel" into a brief an editor can hit. More on that in how to brief your video editor.
Sources: how this was measured
The measurements were taken in August and September 2026 on the public videos linked above, as part of matching reference styles for documentary edits. For Librarian, the stretch measured is 0:00–30:34; for The Interpreter, the first 600 seconds of each of the two uploads.
- Picture changes: frames were sampled from each video and run through scene-change detection, which flags moments where the image changes sharply. Those candidates were then reviewed. "Full-frame" changes count only moments where the whole frame is replaced.
- Transitions: each change was checked frame by frame. If the change happens in one frame, it is a hard cut; if it spreads over several frames, it is a dissolve, a dip to black or a flash.
- A correction on Philip Thompson: an early automatic classifier put dissolves at 48% of the transitions. That was a method error. Eight of the 52 disputed events were checked frame by frame, and in all eight between 95% and 113% of the change happened in a single frame, which makes them hard cuts. The corrected mix is the one on this page: about a third cuts, a third dips to black and just under a third dissolves.
- Text on screen and loudness: text is timed from when a caption or card appears to when it leaves. Loudness is integrated loudness in LUFS, with loudness range and true peak where listed.
Exact numbers move a little with the method, which is why some rows show ranges. What stays the same is the shape: Librarian works with roughly 30-second shots joined by dissolves, The Interpreter cuts hard every 4 to 5 seconds, and Philip Thompson sits in between on shot length but mixes all three kinds of transition. For how the same measurements feed a real documentary edit, see the documentary YouTube channels page.
Pacing FAQ
Is there an ideal number of cuts per minute for YouTube?
Not one that holds across formats. In the three documentary channels measured here, the rate ran from about 2 picture changes a minute to about 13-15, and all three are established channels. Measure the videos your audience already watches in your niche and start from there.
Should documentary videos use dissolves or hard cuts?
Either, as long as it is consistent with the style. Librarian joins 63 of 66 shots with a cross-dissolve of about 2 seconds, The Interpreter used hard cuts in all 26 changes checked frame by frame, and Philip Thompson splits roughly into thirds: about 33% cuts, 28% dissolves and 36% dips to black.
Does pacing stay the same through the whole video?
Not in the Philip Thompson video measured here. The cold open ran at about 19 changes a minute, the first 10 minutes at 14.5, minutes 10-25 at 12.5, minutes 25-30 at 10.8 and the closing at 8.2. The pace eases off as the story settles.