How long should an explainer video be?

Sixty to ninety seconds, for most single-concept product explainers. Across sixty-odd produced videos, the animated core of each finished piece ran 54 to 124 seconds, with the typical video landing at 60–80 — and the ones that ran long earned it with a worked example, never with a longer pitch.

But the more useful answer is a method, not a number: length is an output. You cut the concept into scenes, each scene carries one idea and a minimum duration the picture needs, and the runtime is the sum. Pick a duration first and fill it, and the filling is what viewers feel. Here's the data, the math, and the decision.

The real numbers

From production records, not estimates:

| Format | Measured length | | --- | --- | | Single-concept explainer (the core format) | 54–124s, typically 60–80s | | On-camera founder intro | 25–30s ≈ 70–80 words | | Full product course (11 modules) | ~11 minutes of animation total | | Onboarding sequence | 10–15 minutes across 4–5 short videos |

Two things stand out in that table. First, even a complete course of a serious product is about eleven minutes of animation — a minute per concept. Second, when more total time is needed, it ships as more short videos, never as one long one. A 12-minute explainer doesn't appear anywhere in the corpus because nothing a single video argues takes twelve minutes to argue.

The words-per-second math

The math that keeps scripts honest: a production narration voice reads at about 1.9 words per second, and every scene's line is followed by roughly 0.7 seconds of breath so the read doesn't feel wall-to-wall. Work that through:

• A 9-second scene holds one short sentence — about 15–18 words.
• A 12-second scene holds two.
• A 60-second video holds roughly 100–110 words of narration, total.
• A typical 6–8 scene video at 8–12 seconds per scene comes out at 60–90 seconds — which is exactly where the measured corpus sits.

This is the fastest way to audit a script you've been handed. A 500-word script is a five-minute video. If the target is 90 seconds, that script is overwritten by a factor of three, and no amount of fast reading fixes it — compressing the read produces the amputated-fragment problem, where prose stops being speech. The fix is cutting ideas, never cutting verbs.

The 0.7-second breath matters more than it looks. Narration that runs wall-to-wall reads as anxious; the pad is what makes a machine-produced read feel like a person explaining something. Budget it per scene, not per video.

Length follows scene count, never the reverse

The planning chain that produces every video in the corpus runs: learning outcome → one thesis sentence → causal chain → beats (the chain cut into scenes, one idea each) → visuals per beat → narration last. Duration never appears as an input. It emerges at the beat step: this concept needs six links in its causal chain, each link needs 8–12 seconds of picture, so this is a 60–75 second video.

Two production rules enforce the direction:

• Every scene has a visual minimum — the duration its animation was designed for. A scene's final length is the larger of that minimum and its narration plus the breath pad. Words may extend a scene; they can never compress it below what the picture needs. The picture's timing is a floor.
• If narration keeps running far shorter than a scene's visuals, the visual minimum is padded — the fix is trimming the scene, never padding the prose to fill it.

The consequence: a video is exactly as long as its idea, measured in scenes. When a draft feels long, the cut list is scenes, and the test is the thesis sentence — everything on screen exists to make one idea feel obvious, and any scene that doesn't push it gets deleted no matter how good it looks. That's also why one idea per scene is the load-bearing rule of the whole method: it's what makes length a computable quantity instead of a taste argument.

When 60 seconds, when 2 minutes

The honest fork, from the intake question we ask before scripting: the target length decides whether you get a worked example or only the concept.

60–90 seconds — one concept. Six to eight scenes: hook, mechanism shown macro to micro, one run of the machine, closer. This covers a homepage video, a feature announcement, a "what is this product" piece. Most products need exactly this and nothing longer.

2 to 2.5 minutes — concept plus a worked example. The extra time buys one thing: walking a single real case through the machine mechanistically — a failed run debugged backwards, one document followed from upload to answer. It does not buy a second pitch, more features, or a longer intro. If the added minute isn't a worked example, it's padding.

Longer than 2.5 minutes — you have two videos. Split along concept ownership: each video argues one thesis and defers its neighbors. Viewers arrive non-linearly anyway, and three tight videos get watched to the end where one 6-minute video gets abandoned at the 90-second mark.

Two findings from the graded corpus explain why the ceiling is real. First, run economy: viewers pay full attention to a machine's first run, less to the second, almost none to the fifth — one graded build went from 7 runs to 3 and got stronger. Extra runtime is spent where attention no longer is. Second, padding is visible: one measured bad build spent 44% of its runtime fully static — 52 seconds of settled frames in a 93-second video — and that's precisely what a video stretched past its scene count looks like. The hook earns the first five seconds; nothing earns second 200.

What the length decision actually costs

Runtime scales cost everywhere — more scenes, more narration, more revision surface — but the expensive mistake runs the other way: paying for 2.5 minutes when the idea was 70 seconds. Before commissioning to a target length, write the thesis sentence and count the links in its causal chain. Six links is a 60–80 second video regardless of what the brief says. Pricing implications are on the cost guide; how long production takes is a separate question from how long the video runs.

FAQ

How many words is a 60-second explainer script? About 100–110 words of spoken narration, at ~1.9 words per second with a 0.7-second breath after each scene's line. If your draft is 250 words, you've written a 2-minute video.

Is a 3-minute explainer too long? For a single video arguing a single idea, yes, in our experience. Past about 2.5 minutes you're holding attention you no longer have. Split it: each piece owns one thesis, and the set becomes a sequence viewers can enter anywhere.

Should the ad cut and the homepage video be different lengths? Yes — they're different scene counts, not a re-trim. An ad cut is the hook and the single strongest run, 15–30 seconds; the homepage video carries the full causal chain. Cutting a 90-second video down by playing it faster or dropping its middle breaks the chain that made it work.

Do silent scenes count against the word budget? They don't need words at all — a scene with no narration keeps its authored duration, and some of the strongest beats in the corpus are silent payoffs. Budget words per scene, and let the picture carry the scenes that can.

Not sure whether your product's idea is a 60-second chain or a 2-minute one? Submit your URL and see twenty candidate directions cut both ways within 24 hours.