In your paper, every claim sits next to its confidence interval, and the limitations paragraph does its quiet work. In a three-minute video, the caveats are usually the first thing cut, because they feel slow.
That matters more in medicine than almost anywhere else. A video about a health finding can be watched by a patient, a carer or a journalist, and some of them will act on what they hear. The job is to make the video clear without making it stronger than the paper.
What Medical Journals Ask For
Journals that take video abstracts set their own rules, so start with yours. As one example, here is what BMJ's author hub says at the time of writing (October 2026):
- Timing. BMJ will not ask you to consider a video abstract until your paper has been accepted, so do not upload one at initial submission.
- Review. Every video abstract is assessed for suitability by BMJ's multimedia and editorial teams, publication is not guaranteed, and you may be asked to make small changes.
- Length. The video should be as short as possible and no more than 4 minutes.
- Format. BMJ's technical specs ask for 16:9 widescreen at a minimum of 25 frames per second, with .mp4 preferred and the H.264 video codec.
- Disclaimer. For more sensitive subjects, BMJ adds a note to the video description saying it describes a scientific research study and its contents may not be appropriate for clinical purposes.
Other publishers frame length differently. Taylor & Francis, for instance, says 2 minutes 20 seconds or less is optimal for social media. For the general production steps, see how to make a video abstract.
Where Overclaiming Creeps In
A study of press releases, not videos, maps the traps a script can fall into. A 2014 retrospective observational study in The BMJ looked at 462 press releases on biomedical and health research from 20 leading UK universities in 2011, and measured each kind of exaggeration only in the releases where it could occur:
- Of the 213 press releases where advice appeared in the paper, the release or the news, 40% exaggerated the advice.
- Of the 182 reporting correlational results, 33% made exaggerated causal claims.
- Of the 105 about studies in animals, cells or simulations, 36% exaggerated the inference to humans.
When a press release exaggerated, the news stories about it were much more likely to exaggerate too. The authors also found little evidence that exaggeration increased news uptake. In other words, overstating the finding did not reliably buy more coverage.
Those three categories make a useful script check: advice the paper does not give, causation from a correlational design, and people from a study in animals.
A worked example: rewriting three lines
Say you are scripting videos for three hypothetical papers. Here is how a tempting line compares with one that stays true to the paper:
| What the paper found | Tempting video line | Accurate video line |
|---|---|---|
| In a cohort study, people with higher daily step counts had fewer hospital admissions | "Walking more keeps you out of hospital." | "People who walked more were admitted to hospital less often. This kind of study cannot show that walking caused it." |
| In mice, a compound slowed tumour growth | "A new compound shrinks tumours." | "In mice, the compound slowed tumour growth. It has not been tested in people yet." |
| A pilot trial suggests a full trial is worth running | "Patients should ask their doctor about this treatment." | "These results need a full trial before anyone changes treatment." |
A simple rule: if a sentence would not survive peer review in your abstract, it should not be said on camera.
Numbers Patients Can Follow
If any version of your video is meant for patients or the public, say numbers the way the Patient Information Forum's guide on communicating benefits, risks and uncertainties recommends: natural frequencies such as 10 in 100, comparisons out of the same number, and absolute rather than relative risk on its own. On screen, that usually means one absolute number per result, written out, with the denominator visible. For the full set of rules and how they apply to clinicians as well as patients, see sharing clinical research with patients.
Patients on Camera and Patient-Facing Versions
Clinical footage and images need the same care in a video as in the paper. BMJ's patient consent and confidentiality policy says that if there is any chance a patient could be identified from a photograph or other static or moving image, or from the text that accompanies it, BMJ needs the patient's written consent. Images such as x rays, pathology slides or images of undistinctive parts of the body may be used without consent if they are anonymised, with identifying marks removed and no accompanying detail that could reveal who the patient is. Check your own journal's policy and your ethics approval as well.
A patient-facing version is best treated as a separate cut, not a softer edit of the scientific one. BMJ's top tips are a useful starting point even though BMJ's video abstracts are aimed at the scientific public: write the script as you would say it, imagine telling the story to a lay person, and remember that a more accessible video reaches a wider audience, including journalists. For a patient audience, lead with what the study asked and who took part, then the result in absolute numbers, then what it does not mean yet.
A final check before you publish
- Every spoken claim matches a sentence in your paper.
- The study design is named out loud: trial, cohort, case series or animal study.
- Numbers are absolute, with the same denominator for comparisons.
- At least one limitation is said, not just written in the description.
- Every patient image has consent or is properly anonymised.
- The journal's length, format and AI policy have been checked.
Where Amplify Fits
Amplify makes a video abstract (16:9, about 3 minutes) from the PDF you upload and nothing else, with no open-web filler. Every claim is traced to a page in the PDF, and a subject-matter reviewer reads every draft before it can leave your workspace.
The video arrives as three variants that you can edit line by line or regenerate, which is where you catch a sentence drifting toward advice or causation. Nothing publishes until you approve it. The same order also includes a vertical short (9:16, 45 to 60 seconds), an audio brief in English plus 21 more languages on request, and a one-page, print-ready infographic, each carrying the title, the authors and a link back to the paper of record.
Whether a journal hosts it is the journal's decision. If you plan to submit it, check the journal's specs and AI policy first, and ask the editor if unsure.
A careful medical video does not need to be dull. It needs the study design said plainly, numbers people can follow, and nothing the paper itself would not stand behind.
