How long should a work video be? Lengths by purpose

No single number fits every recording, but each kind of work video has a sensible default. Start from the ask, aim for the low end, and give the rare twenty minute video chapters so nobody has to watch all of it.

By the VeoRec team · Oct 3, 2026 · 12 min read

How long should a work video be? Shorter than the one you were about to record, most of the time. A quick update turns into four minutes because you add context, then a caveat, then a second thing you remembered while talking. Nobody planned a four minute video; it just grew.

The fix is to decide the length before you record, the same way you would decide how long a meeting gets. Below are the defaults worth starting from for each kind of work video, what one large study of video viewing found about attention, and how to handle the handoffs and onboarding sessions that really do need fifteen minutes or more.

Use these default lengths by purpose

These are working defaults from experience, not figures from a study. Each one is set by two questions: how much the viewer has to see, and how likely they are to watch now rather than "later", which often means never.

The pattern across the table: the more a video has to do (several screens, several decisions, an unfamiliar audience), the longer it can be. The more often it will be watched, the shorter each piece should be.

Why the most frequent videos should be the shortest

Status updates and bug reports sit at the short end for a reason that has little to do with attention spans. They are frequent. A team that sends five updates a day and twenty bug videos a week builds a habit around them, and the habit is either "watch them as they arrive" or "let them pile up". Two minute videos tend to get the first habit and six minute videos the second. Once a queue of unwatched recordings exists, people start skipping all of them, including the urgent one.

Rarer videos can afford more. A quarterly roadmap walkthrough or a recorded client review happens once, the audience expects it, and they will set time aside. Length there is less of a problem than structure, which is covered further down.

What the research on video length actually says

Large public studies of internal work videos are hard to find. The numbers that circulate come from online courses and marketing, and they get garbled as they spread ("people only watch for six minutes"). The originals are more useful than the rumours, so here they are as the sources state them.

The most cited one comes from edX. In How Video Production Affects Student Engagement (Guo, Kim and Rubin, 2014), the authors analysed 6.9 million video watching sessions across four courses. Their findings on length, measured on sessions played at normal speed:

• Median viewing time was at most about six minutes, no matter how long the video was. Making a video twelve minutes long did not buy twelve minutes of attention; the typical session still ended at six minutes or less.
• For videos longer than nine minutes, viewers often made it less than halfway through.
• The shortest videos, up to three minutes, had the highest engagement: in three quarters of those sessions, people watched more than three quarters of the video.
• People were less likely to attempt the exercise after a longer video. The share of sessions followed by a problem attempt fell from 56% for the shortest group to 31% for the longest.

Normal-speed plays were 76% of all sessions. The rest used the other speeds the player offered (0.75x, 1.25x, 1.5x) or switched between them, and the authors note that people playing at 1.25x and 1.5x finished videos faster. Expect colleagues to do the same with yours: a video that feels a reasonable length at 1x is being skimmed at 1.5x by people hunting for the part that concerns them.

On the marketing side, the video host Wistia publishes engagement figures. At the time of writing, its guide to video length reports that videos under a minute average a 52% engagement rate, and that viewers typically watch more than half of educational and tutorial videos between one and five minutes long. Different audience, same lesson: attention is easiest to hold for the first few minutes.

Measure length in viewer minutes, not video minutes

The number on the player is the wrong one to look at. Multiply it by the number of people who will watch. A six minute walkthrough sent to one reviewer costs six minutes. The same video posted to a channel of fifteen people costs ninety if they all watch, which is a meeting.

This matters most for recordings that go to big groups: weekly team updates, all-hands summaries, release walkthroughs. A two minute cut from a six minute original saves sixty minutes across fifteen viewers, and it probably took you ten minutes to make.

It also helps to compare with the written version. Most people get through a page of text quicker than they could hear it read aloud, so a video has to earn its extra time by showing something text cannot: a flow, a bug happening, a tone of voice. Put in your own word count and reading speed here:

If the video version wins only because you did not want to write, send the text. If it wins because the viewer must see something move, record, and keep it short. Our guide on choosing between video, text and a meeting covers that decision in more detail.

One video, one ask

The clearest warning sign of a too-long work video is a second ask. "Can you approve the layout, and also, while I have you, the copy needs a look, and I wanted to flag the analytics thing" is three videos sharing one link.

Multiple asks hurt in two ways. The video gets longer, and the later requests get lost: people act on the first one and stop watching, or remember the last one and forget the first. If different people own the different asks, each of them watches the whole thing to find their part.

So before recording, write the ask down. If you find yourself writing "and", decide whether the two belong together. A layout and its copy for the same reviewer might. A layout for the designer and an analytics question for the data team do not. Our guide to scripting a screen recording walks through building a recording around a single ask.

Front-load the message so any length works

Whatever length you settle on, assume the viewer may stop at any point. Some will watch to the end. Others will get the gist in twenty seconds, decide it is not urgent and close the tab. The edX numbers say that is normal behaviour, not rudeness. So arrange the video so that stopping early still leaves the viewer with the most important thing.

In practice, that means the order is inverted from how most people tell a story. Do not build up to the point; start with it. A useful structure for most work videos:

1. First ten seconds: what this is about and what you need, with the deadline. "The export bug is fixed on staging; I need a yes to ship it on Thursday."
2. Next minute: the evidence or the walkthrough that supports it. Show, do not narrate the history.
3. After that: detail for the people who want it, such as edge cases, alternatives you rejected, what you would do next.
4. Last ten seconds: repeat the ask in one sentence.

A viewer who stops after the first ten seconds knows what you need. One who stops after a minute has seen the proof. Only the people who care about the edge cases sit through the edge cases. The video can be as long as it needs to be, because nobody is forced to watch past the part that concerns them.

Put the same ask in the message that carries the link: "Export fix ready, need a yes by Thursday (1:40)". Including the duration helps people decide when to watch, and a short number gets watched sooner.

Spot a video that is too long before you send it

You can usually tell from the recording itself. Watch for these:

• The first fifteen seconds have no information. Greetings, "so", "let me just share my screen", a tab search.
• You say "as I mentioned". You are repeating yourself, probably because you lost your place.
• There is silent clicking. Pages loading, menus you opened by mistake, a search for the right tab.
• You show everything you built, not what the viewer needs. The reviewer needs the changed screen, not a tour of the app.
• The ending trails off. "So, yeah, I think that is it, let me know" with no clear ask.
• You covered a topic the viewer did not ask about. Interesting to you, a detour for them.

Each of these is fixable in a minute or two of editing, which is usually quicker than re-recording.

One more test that works well: describe the video to yourself in a single sentence, the way you would in the message that carries the link. If the sentence needs "and also", or if you cannot finish it without going back to the recording to remember what you covered, the video is doing too much. Shorten it, or split it into two links with two clear subjects. A recording you cannot summarise is one the viewer will not be able to act on either.

Cut a long recording down in five minutes

Most recordings can lose a fifth of their length without losing anything the viewer needs. Work in this order, because the first steps are the cheapest:

1. Trim the start Cut everything before your first useful sentence. This alone often removes ten to twenty seconds and makes the hook the first thing heard.
2. Trim the end End right after the ask. Everything after "let me know by Thursday" is the viewer waiting for nothing.
3. Remove silences Page loads, pauses while you think, the gap after a fluffed sentence. Automatic silence removal handles most of them.
4. Cut detours A tangent, a repeated explanation, the wrong tab you opened. Cut the section out and let the two ends join.
5. Split if there are two asks If the middle is really a second video, split it there and send two links with two clear subjects.

A worked example: a 6:40 update cut to 2:20

Take a weekly update recorded for a team of twelve. The first take ran 6:40. Here is where the time went, and what happened to each part:

What is left is 1:55 of shipped, next and blocked, plus about 25 seconds for the one-sentence reason and the ask: 2:20 in all. The 4:20 saved, times twelve viewers, is 52 minutes of the team's attention, for a few minutes of editing. And the data team gets a 1:20 video that is entirely about them, instead of having to find their part in the middle of someone else's update.

All of this can be done in the browser: VeoRec's editor trims, splits, cuts and removes silences. Our guide to editing screen recordings shows each step.

When a long video is the right call

Some recordings are long because the job is long. A handoff of a project you have run for a year, an onboarding walkthrough of a codebase, a recorded review of a full website for a client. Squeezing those into two minutes just moves the work to a follow-up meeting.

The research suggests what to do about it. In the edX data, viewers of tutorials watched only two to three minutes of each video on average, whatever its length, and came back to rewatch tutorials more often than lectures. The authors' reading was that people re-watch long tutorials and jump to the parts they need, so they recommended signposts that make skimming easy. A long work video gets used the same way, which means it has to be easy to jump around in.

That means three things:

• Chapters, so viewers can see the structure and skip to "Deploying" or "Known issues" without scrubbing.
• A short summary at the top, in the message or description, so someone who only needs one fact can get it without pressing play.
• A searchable transcript, so "where did she explain the cron job?" is a search, not a rewatch.

A fourteen minute handoff with chapters looks like this. Click one to jump:

A viewer who only needs the reconciliation job watches three minutes, not fourteen. In VeoRec you add these as you record by pressing M at each topic change, then name them on the finish screen; on Pro, AI chapters are generated after each recording. Every recording also gets an automatic transcript you can search across your library. Our guide to chapter markers covers naming chapters people actually click.

Split a long topic into a short series

Chapters work for one long recording. Often a series of short ones works better, especially when different people need different parts, or when parts will go out of date at different speeds.

Split by what the viewer is trying to do, not by how you happened to explain it. For an onboarding set, that might be: set up the environment, run the tests, ship a small change, where things live in production. Each one stands alone: it starts by saying what you will be able to do by the end, and it does not assume the viewer watched the one before.

Name them so the order and topic are obvious from the title alone: "Onboarding 1 of 4: local setup (6 min)". Put the series in one folder and share the folder rather than four separate links. When the deploy process changes, you re-record part four and leave the rest alone.

Short parts also fit recording limits more comfortably. VeoRec's Free plan records up to ten minutes per video, which turns out to be a useful forcing function: if you hit the limit on a status update, the update was too long. Pro allows recordings up to ten hours for the occasional genuine marathon, such as a recorded workshop. Plan details are on the pricing page.

Test your instincts on four real cases

A quick self-check. For each situation, pick the length and format you would use.

What to do with your next recording

Before recording, write the ask and look up its default in the table above. Aim for the low end. If you go over, do not re-record: trim the start, trim the end, remove silences, and check whether the middle is a second video.

For the long ones, accept that they are long and make them easy to move around in: chapters at every topic change, a three line summary in the message, a transcript people can search. And look at your last five recordings with fresh eyes. If most of them are over five minutes, the problem is probably not your topics. It is the second and third asks hiding inside them. If those recordings are replacing a recurring meeting, our guide to replacing status meetings with video updates covers formats that stay short week after week.