Research paper to video, without losing the nuance

What to cut, in order

A twenty-page paper is roughly ninety minutes of narration. The video is six. So this is almost entirely a cutting exercise, and papers have a conventional structure that makes the order obvious.

• Related work — cut entirely. It exists for reviewers.
• Methodology — reduce to one sentence naming the design and the sample. Anyone who needs more will read the paper.
• Robustness checks and appendices — cut, with a line saying they exist.
• Limitations — keep. See below; this is the one people wrongly cut.
• Finding, its strongest figure, and the implication — this is the video.

The hedges that are load-bearing

Academic writing hedges deliberately, and compression removes hedges first because they add words without adding claims. That is exactly backwards. 'Associated with' becoming 'causes', 'in this sample' becoming a general statement, 'suggests' becoming 'shows' — each is a small edit and a substantive misrepresentation.

The rule that works: any word doing epistemic work stays, even when it costs runtime. If the narration cannot fit the qualifier, the sentence is wrong at that length and needs restructuring rather than trimming.

This is also why the narration script must be read against the paper before rendering. An automated rewrite will produce fluent, confident sentences, and confidence is precisely the failure mode here.

Keep the limitations section

Counter-intuitive, and it is the thing that most distinguishes a credible research video from a press-release one. Thirty seconds acknowledging what the study cannot show buys more trust from a knowledgeable audience than two minutes of additional findings.

It also protects you. A paper video that over-claims gets quote-tweeted by people who read the paper, and that correction travels further than the video did.

Figures, at readable size

Your figures are the reason a research video beats a talking head, and they are almost always unusable as exported. A figure designed for a two-column journal layout has 7pt axis labels that become invisible at phone size.

Crop to the panel that carries the finding — usually one of four — and increase the label size. Then let the narration point at the specific feature rather than describing the chart in general: 'the gap opens after week six' beats 'this chart shows the treatment effect over time'.

• One figure per scene, never two.
• Crop to the panel that matters.
• Axis labels at least as large as the body text.
• Narration points at a feature, not at the chart.

The structure that works

Finding first. Academic writing builds to the result; video loses the audience before it arrives. Open with what you found, then spend the middle on why it is credible, and close with what it changes.

This inverts the paper deliberately, and it is the single biggest structural change. Everything else — the cutting, the hedging, the figures — is easier once the finding is at the front.

Frequently Asked Questions