A teleprompter for video abstracts — the terminology, said exactly

Journals increasingly ask for a video abstract alongside the accepted manuscript, and conferences have moved half their programmes to prerecorded talks. For a researcher this is writing in a medium with none of writing’s safety. The wording of a paper was negotiated — with co-authors, with reviewer 2 — and every hedge is load-bearing: ‘in this cohort’ and ‘consistent with’ are not filler. Improvise on camera and the claim quietly grows past what the data supports.

The terminology is its own hazard. Dense terms have to be said exactly, once, on camera — phosphatidylinositol does not forgive a fumble, and neither does an audience that uses the word daily. Reading fixes that, but reading from a document beside the camera is visible in every frame of a talk aimed at the sharpest audience you will ever have.

Captions raise the stakes again. Much of the audience meets the video with the sound off, accessibility requirements increasingly make captions mandatory, and a good part of the field reads English better than it hears it. Auto-generated captions mangle exactly the words that matter — the enzyme, the assay, the acronym — and a caption that renders CRISPR-Cas9 as ‘crisper cast nine’ is not a typo, it is a credibility problem published under your name.

And the visual is usually missing. The interesting thing — the apparatus, the poster, the specimen, the field site — sits behind you while you narrate slides, because showing it would mean a second camera, a second clip and an editor. So the video abstract shows a face and a slide, when the point of the medium was to show the thing itself.

The default workflow is narrating a slide deck in the conference’s recording tool or a screen recorder, in one anxious take per attempt — a fumbled term at minute four means starting minute one again, because the tool has no way of picking up mid-take.

Captions are handled after the fact: run auto-transcription, then spend an hour walking its guesses line by line, correcting every technical term and every author name, for every version of the talk. Some labs pay a captioning service and proofread it anyway.

The apparatus shot, when it happens at all, is a separate clip filmed on a phone and stitched to the talk by whoever in the lab knows an editor — which is how a three-minute video abstract comes to cost an afternoon.

Prerecorded conference talks add a duration cap enforced by the upload portal — twelve minutes, strictly — so the whole cycle of fumble, restart and re-caption runs against a countdown, usually in the same week the camera-ready paper is due.

The script is the abstract you already wrote, adapted for speech, and the prompter holds you to it. Voice follow scrolls the text as you speak and waits through the deliberate pause before the key result; skip the sentence you decided to cut and it catches up. Recognition runs on the device and nothing you say is uploaded. The wording your co-authors signed off is the wording that gets said — hedges intact, terms exact.

Captions come from that script, not from a recognition engine’s guess at it. The caption text is your text — phosphatidylinositol spelled the way the journal spells it. The lines are timed against your actual delivery, then export as a separate .srt file, which is what journal portals and conference platforms ask for.

Showing the thing no longer needs an editor. Mid-sentence, you flip from the front camera to the rear one and the recording carries on — one continuous file, no cut, no second clip to stitch, because the encoder never stops. Narrate to the lens, flip to the poster or the rig while you keep talking, flip back for the conclusion.

And the take remembers its script. The text you read is saved beside the recording, so when the revised manuscript comes back months later and the video needs one changed sentence, you open the old take and the exact narration is sitting next to it — ready to correct, re-record and re-caption without reconstructing anything from the video.

None of it needs a second pair of hands or a booking in the media suite. The recording, the flip, the captions and the .srt all happen on the phone — which, for a lab, turns the video abstract from a favour someone owes the PI into an hour in the calendar.

A materials scientist has a paper accepted and a four-minute video abstract to deliver with it. She adapts the abstract and key findings into a script, records to the front camera at the bench, and at ‘the cell we built for this measurement’ flips to the rear camera and keeps reading while the rig fills the frame. Flip back, conclusion to the lens, stop. One file, nothing to stitch.

Captions export as an .srt in minutes, and her co-author checks them against the manuscript — they match, because they are the manuscript. When the journal’s production team later asks for one sentence to change, the take still carries its script: she edits the sentence, re-records the closing section, and the new captions are as exact as the old ones. Total cost of the revision: twenty minutes, including finding the phone.