Skip to content
Primores AI wiki

Turning an Article into a Short Explainer Video: Formats, Scripting Rules, Evidence Boards

Turning an Article into a Short Explainer Video

TL;DR: Pick the video format from the article’s shape (a numbered list, one number, a how-to, a myth), not from habit. Script at most three ideas per 45 seconds, make the hook readable on the very first frame, and budget for the pace explainers actually run at: six measured TikTok explainers ran 178-233 words per minute, about the speed of ordinary conversation. In that small set, the how-to clips earned 4-8% saves and the 6.8M-play viral one under 1%. That suggests explainers should be built to be saved, not to go viral. It is a hypothesis from six videos, not a benchmark.

If you want this run as an ongoing programme, Primores works as a brand’s TikTok content team: carousels and videos produced, formats and angles tested, the winners scaled: TikTok content.

How solid is this page? It comes from building an article-to-video tool and studying six reference TikToks picked by taste, on one platform, in November 2026. The format menu and the scripting rules are practitioner rules unless a source is cited next to them. The evidence board, the save-rate observation and the series pattern are hypotheses until Primores has completion and save data from its own posts.

Pick the format from the article’s shape

The format is a separate choice from the visual style. Most articles fit two or three formats; rotate them so a feed doesn’t repeat itself.

FormatFits articles that are…What it looks like
Character storyan argument or how-to with a clear before → afterAbstract things become characters; one visual surprise per beat
Kinetic quoteone strong idea or principleSlow abstract motion, 1-3 keywords, the sentence word by word; 15-35 s
Swipe carousellists, frameworks, steps, checklists6-10 stills, one idea per slide; see marketing/slideshow-pattern-design for the nine slide patterns
Data storystats, benchmarks, before/after numbersOne number per beat, counters and bar reveals, the source on screen
Screen demo”do this in the tool” tipsA rebuilt app or browser UI, typing, a highlight on the result
Chat / thread storyobjections, myths, customer questionsMessage bubbles in sequence; the answer lands in the last one
Myth vs factcorrecting a common mistakeSplit screen, “what most do” vs “what works”, three pairs at most
Whiteboardframeworks, processesLines drawing on, a marker cursor
Quizsurprising factsQuestion, three options, countdown, reveal
Talking head + layerthe author wants to be on cameraPresenter cut against full-screen cards (the references cut, they don’t overlay)
Reaction / fact-checksomeone else’s viral claimThe original clip, a pause at each claim, a sourced verdict
Evidence boarda numbered list of risks, mistakes or reasons, each with a sourceVoice-led; every claim appears as its source card (below)
Step timelinesetup guides, numbered proceduresOne long canvas, numbered nodes, the camera glides from step to step
Pictogram tapeone surprising finding or mechanismFlat stick figures, a tape label on every noun; built for shares, not saves
Tool walkthrough”someone built X, it’s free”The product’s UI rebuilt as clean cards, a presenter or voice-only

Quick routing: a sourced listicle → evidence board; setup steps → step timeline; one big idea → kinetic quote; numbers → data story; “most people think…” → myth vs fact or chat thread; a list for later reference → carousel, and also a video version.

Carousels are their own discipline. The nine slideshow patterns, which signal each one tends to earn (saves, shares, comments, follows) and why, are on marketing/slideshow-pattern-design. This page doesn’t repeat them; it adds the video formats around them.

Scripting rules, with numbers

RuleStatus
At most 3 ideas per 45 s. Cut the rest; the article carries the detail.Practitioner rule
The hook is fully readable on frame 0. Frame 0 is the thumbnail and the first thing a scroller sees.Practitioner rule. TikTok’s own guidance is looser: “Prioritize your hook in the first 6 seconds” and “introduce your content proposition in the first 3 seconds”
On-screen text at a readable rate.TikTok recommends “5-10 words per second when using text”
Pace: budget ~180 words per 60 s for a voiced explainer.Measured: the six references ran 178, 183, 192, 205 and 233 wpm (one was music only). An early budget of ~120 words per 60 s came from self-recorded reads, which run 30-50% long; it is a recording budget, not the pace of a finished explainer
A new visual every 3-6 s, or one continuous canvas with a camera that never stops movingObserved in the six references (the top saver had 0 cuts in 81 s). TikTok recommends transitions and graphics “to retain engagement and view time” but gives no interval
Captions: one word or short phrase at a time, key word boxed or colouredObserved in all six references

Why ~180 wpm isn’t “fast”. In 2,438 recorded English phone conversations (the Switchboard corpus), people averaged 196 words per minute, with individual conversations from 111 to 291 (Yuan, Liberman and Cieri, Interspeech 2006). A voiced explainer at 180-200 wpm runs at conversational speed. The 120 wpm feel of a nervous self-recording is the outlier.

The evidence-board pattern (hypothesis)

A voice-led listicle in which every claim appears on screen as its source. Structure, for about 60 seconds and ~180 words:

  1. Hook: one line with a surprising number (“your app can be fined $X before it earns a cent”), then “here are N reasons why”.
  2. Each item, same three steps: a situation (a 12-year-old signs up), a mechanism (your form never asked their age), and a penalty with its source on screen (the statute or the regulator’s page, quoted briefly).
  3. Recap that ties the items into one line, a payoff, the CTA.

On screen: a headline band below the top safe zone carrying the current spoken phrase, source cards on a paper-like board with the cited line marked, numbers stamped as they’re spoken, and the camera moving card to card every 1.5-3 s.

Why it might work: it is glossary/honest-assessment and glossary/geo-anchor turned into video. Sources are visible, numbers are stamped with their origin, and the viewer can check every claim, which is also why it’s worth saving. The reason to hold it loosely: it was singled out from one external TikTok that performed well. Primores has no completion or save data on its own evidence-board videos yet.

Save rate runs against plays (hypothesis)

In the six references, measured in November 2026:

Reference typePlaysSave rate
Psychology “fun fact” (pictogram)6.8M0.99%
Step-by-step setup guide143k7.9%
Tool walkthrough82k5.4%
Presenter + full-screen cards (method + plan)23k5.2%
Launch news16k4.0%
AI film-look mood clip (not an explainer)46k0.73%

The pattern: plays measure reach, which the hook and the topic drive. Saves mean “I’ll use this later”, which is what an explainer is for. So optimise explainers for saves, and don’t copy the formats of viral entertainment. That matches the save-bait column of marketing/slideshow-pattern-design (step-by-step and numbered lists earn saves; fact-stacks earn shares) and the save ratio in marketing/behavioral-profile-fingerprinting.

Limits: n = 6, chosen by taste, one platform, one point in time. It is a direction for what to test, not a benchmark to report.

A series from one article (hypothesis)

One article usually carries more than one video. Instead of cramming a long article into 60 seconds against the three-ideas rule, plan a short series:

  1. Name one pain the article solves, in the reader’s words. If there isn’t a clear one, make one video.
  2. Pick 5 angles, 10 at most, each built only from the article’s own content: the fear (what goes wrong), the argument (what most believe and why it’s wrong), the do-it-now (the steps), the number (the one surprising stat), the story (a before → after), the tool, the quiz.
  3. Alternate formats so no two neighbouring videos look the same.
  4. Lead with the strongest hook; put the do-it-now video second, since it is the one most likely to be saved.

This is marketing/discovery-before-scale at the scale of one article: several patterns tested against the same audience and the same pain, cheaply, before deciding which format to scale. It has not been run end-to-end yet.

Key Takeaways

  • Choose the explainer format from the article’s shape; carousels follow marketing/slideshow-pattern-design, the video formats are the table above.
  • Script at most three ideas per 45 s, make the hook readable on frame 0, and budget ~180 words per minute: that’s conversational speed (Switchboard averages 196 wpm), not fast.
  • The evidence board (every claim shown as its source card) is honest assessment in video form. Treat it as a hypothesis; it rests on one external example.
  • In six references, how-to explainers saved 4-8% while a 6.8M-play viral clip saved under 1%. Aim explainers at saves. Six videos is a hypothesis.
  • One article can feed a 5-10 video series with alternating formats: a low-volume discovery test of which format works.
  • Keep the layout inside each platform’s safe zone: marketing/short-form-video-safe-zones.

Sources

By Andrej Ruckij