How do you hook viewers in the first five seconds of an explainer video?

Open cold: on the machine already running, or on the viewer's own problem — never on a logo, a title card, or a warm-up. At a normal narration pace of about 1.9 words per second, the first five seconds hold nine or ten spoken words and one picture, so both have to be doing work before the viewer's thumb decides anything.

Across sixty-odd produced explainers and three graded batch reviews, the opens that survived review fall into exactly two shapes, and the opens that died fall into a handful of named failure modes. This guide covers both, plus the six-step anatomy behind the strongest scripted open we know.

Why the open gets a disproportionate share of the craft

Attention is a budget, and it is highest at second zero. Viewers pay full attention to the first end-to-end run of a machine on screen, less to the second, almost none to the fifth — the same decay applies to the video itself. Whatever you show first is the one thing you can be certain gets watched.

That has a hard consequence: the open cannot be spent on setup. Every second of logo, mission statement, or "in this video we'll cover" is your highest-value attention paying for your lowest-value content. In our review notes, the deadliest kill-word for a whole video is "no machine-at-work" — and it lands hardest when the first framing sits still while a voice clears its throat.

The cold-open doctrine

The rule from production: when a video dies in review for feeling lifeless, the cure is almost always staged at the front. One rejected flagship was diagnosed as "one framing for 100 seconds, polite assembly, no machine-at-work." The restage that passed opened cold on the finished machine running once, with the camera riding the run — alive at the first cut.

Cold open means the first frame is already the subject. Not a build-up to the subject. The workflow is on screen, the run fires, and the voice starts explaining what the viewer is already looking at. Three practical rules make it work:

• Something moves in second one. A pulse leaving a trigger, a row landing, a status flipping. Motion is the promise that this is a demonstration and a demonstration is worth watching.
• The motion is legible. A climax smaller than about 1/40 of the frame didn't happen, and that floor applies double in the open, when the viewer doesn't yet know where to look. Frame the first event large.
• The first line is about this screen. The anti-test we hold every narration line to: if a line could play over any video, it isn't doing its job. That goes triple for line one.

The two opens that pass review

1. The problem-exposition open

Start from the viewer's situation, before the product appears. The reference script in our corpus opens: "You probably already have knowledge that you'd like to integrate into your workflow…" — then walks the obvious solution and its concrete failure ("you'd quickly max out the model's context and get degraded performance") before the product enters as the answer.

This open earns the product's entrance. By the time the capability appears, the viewer has already felt the gap it fills, so the reveal lands as relief instead of a pitch. Use it for concept videos — a new idea the viewer needs a reason to care about.

2. The machine-running open

Start on the product mid-demonstration and orient fast. Another accepted opener, in full: "Here is a real workflow. A trigger, a file parser, an agent, and a database. Watch how a file moves through it." Three moves in two breaths: orient, enumerate what's on screen, issue one watching instruction.

Use it when the capability is the hook — showcase videos, launch videos, anything where seeing the thing run IS the argument. The viewer's question ("what am I looking at?") gets answered in the first sentence, and the second sentence hands them a job: watch this.

Which to pick: if the viewer already wants what you do, run the machine. If they don't know they want it yet, open on their problem. A product demo and an explainer split along the same line.

The hook anatomy, in six steps

The strongest scripted open in our corpus follows the same arc every time. It runs about fifteen to twenty spoken seconds — the five-second hook is its first step plus the right picture — and each step exists for a reason:

1. The you-hook. Start from the viewer's existing situation, with the viewer as the subject. "You probably already have…" Nobody skips a sentence about themselves.
2. The naive path. Walk the obvious solution and name its concrete failure. This is where credibility is built: you demonstrate that you understand the problem at working depth, and the failure has a number or a mechanism attached, never a vague "that doesn't scale."
3. The want. Restate the requirement in plain terms: "What we'd like is a way to search and find only the most relevant pieces." Now the viewer holds a precise, open question.
4. The reveal. The capability enters as the answer to that exact question: "This is where the knowledge base comes in." Because steps two and three built the slot, the product drops into it instead of interrupting.
5. Mechanism. Zoom from macro to micro, defining each term as it appears. The hook has done its job; now the script teaches.
6. The closer. End on behavior or payoff — what the viewer can now do — never a slogan.

The steps are cheap to check: read your open and mark where the reveal lands. If the product name arrives before the viewer holds an open question, steps two and three are missing and the open is a pitch.

The named hook failure modes

Every dead open in our corpus is one of these.

The trailer-voice open. Suspense fragments and personified machinery: a rejected draft opened "A ticket comes in, and a run crosses the chain — triage, build, log. From out here, it's already over." The rejection note: we are making explainers, not marketing. Trailer voice performs instead of explaining — the voice sells while the picture teaches, and they read as two different videos. The accepted rewrite states the screen and names the concept in the first beat: "Here's a workflow processing a support ticket… Every time a workflow runs, the platform records exactly what happened. That record is called a log."

The throat-clear. Logo animation, company mission, "in this video you'll learn how to…" A production rule we hold scripts to: the learning objective is stance, never stated in narration. You design toward it; you never announce it. Announcing the lesson spends the open on a promise the video should simply keep.

The text-wall open. A title card plus on-screen sentences restating the voice. The tell is the mute test: if a viewer could watch the open silent and read the whole setup, the screen is a slide, and slides don't hook. The voice owns the words; the screen owns motion and state.

The invisible open. Something does happen — but at a size or pace nobody registers. A one-word change in a small box on a wide static frame is the same as nothing. The 1/40-of-frame floor exists because a director watched exactly that open and wrote "don't see anything cool."

The count open. "The five ways to…" Never frame by a fixed count; counts date, and the count is never the lesson. Name the capability instead.

Most of these have the same root: the open was written before the picture existed. Which leads to the last rule.

Write the hook last

Counterintuitive but consistent across the corpus: narration is written to finished visuals, never the reverse. The open's key word has to land on the open's key visual moment, and that sync is only achievable in one direction. Decide the concept, lock the scenes, build the picture — then write the nine or ten words that play over second one, the way you'd write a headline.

The rest of the video has the same discipline behind it — one idea per scene, scene count driving length, payoffs staged at frame scale. The open is just where the discipline is most visible, because it's the only part everyone watches.

FAQ

How many words fit in the first five seconds? About nine or ten. Production narration reads at roughly 1.9 words per second, and a breath pad follows every line. Write the opening line like a headline: one subject, one verb, about this screen.

Should the product name appear in the first five seconds? In a machine-running open, yes — orienting fast is the point. In a problem-exposition open, no: the name lands at the reveal, after the naive path has built the slot for it. Naming the product before the viewer holds an open question turns the open into a pitch.

Do question hooks ("Tired of X?") work? The you-hook does the same job better. "You probably already have…" starts from the viewer's real situation and flows into the naive path; "Tired of X?" is an ad pattern viewers have learned to skip, and it commits you to trailer voice.

Is a cold open risky if viewers don't know the product? The open on screen still has to be legible to a stranger — that's what the orienting line is for ("Here is a real workflow. A trigger, a file parser, an agent, and a database."). If the machine can't be oriented in one sentence, open on the problem instead.

Want to see twenty different opens for your own product before writing one? Submit your URL and judge real candidate directions in 24 hours.