01
1. Choose the format the story needs
Each medium carries something the others cannot. Text carries precision, structure and searchability. Stills carry a single moment and are the cheapest form of proof-of-presence. Audio carries voice, hesitation and place. Video carries process and movement. The decision is not which is most engaging but which of those the story actually turns on.
Two practical constraints follow. Every additional format multiplies production time, and an under-produced version damages the story more than its absence would. And every format carries its own accessibility duty, so choosing three formats means committing to a transcript, captions and alt text as well. Decide before you leave the desk, because the capture requirements differ.
02
2. Audio: the four things that matter
Recording usable speech comes down to distance, position, level and room. Get the microphone close, ideally a hand's width from the speaker and off to one side of the mouth to avoid breath. Set the level so peaks sit below the top of the scale, because clipped audio cannot be repaired. Choose the quietest available space and prefer soft furnishings to hard rooms.
Record thirty seconds of room tone at every location, with nobody speaking. It costs nothing and it is what lets an editor cover a cut without an audible jump. Say the date, location and the person's name and role into the recorder before you begin, so the file identifies itself. Where a clip will be published, note the in and out timestamps at the point of capture.
03
3. Vertical video without a crew
Short vertical video is now a normal reporting output and its failures are consistent: shaky framing, inaudible speech, and length beyond what the material supports. Stabilise the phone against something solid, frame with the subject's eyes about a third from the top, keep the light behind you rather than behind the subject, and use an external microphone for any speech that matters.
Structure the clip like a news story: the consequential material in the first three seconds, then the supporting detail, then a plain end. Shoot more than one take of any piece to camera and shoot cutaway material for the same reason a photographer shoots wide and tight. Keep the published clip as long as the material justifies and no longer; padding is more damaging than brevity.
04
4. Stills, cropping and the line at manipulation
News photography has a bright line. Adjusting exposure, contrast and colour balance to represent the scene as it appeared is normal processing. Adding, removing, moving or duplicating content is manipulation and is not publishable as journalism. Cropping is legitimate for composition and illegitimate when it changes what the image shows, such as cropping out the crowd that gave a gesture its meaning.
Keep original files and never overwrite them. Where a portrait was arranged, say so in the caption. Where a photograph is illustrative rather than of the event described, label it. Where an image is supplied by an interested party, name the supplier, because provenance is part of the information the image carries.
05
5. Captions and alt text are different jobs
A caption is a factual statement for all readers: who or what is shown, where, when, and any qualification the reader needs to interpret it. Alt text is a description for people who cannot see the image, and it should convey the information the image contributes rather than repeating the caption word for word. A decorative image gets an empty alt attribute; an informative one needs a real description.
For charts and diagrams, the alt text must carry the finding, not the shape: what the data shows rather than that there is a line going up. Where the image contains text, that text must appear in the alt description or in the surrounding copy, because an image of words is unreadable to a screen reader and to search.
06
6. Subtitles, transcripts and reading order
Every published clip with speech needs a text alternative. For video that means captions synchronised with the speech and identifying who is speaking. For audio it means a transcript, which also makes the material searchable and quotable. Automatic transcription is a starting point that must be corrected by hand, particularly for names, place names, figures and technical terms.
Check three further things before publishing. That any burned-in text has enough contrast against the moving image behind it. That nothing essential is conveyed by colour alone. That the page order still makes sense when the media does not load, since a story that depends entirely on an embedded player has no fallback for a reader on a poor connection.
07
7. Consent, re-use and the second life of a file
Consent to be recorded for one story is not consent to be used in another context indefinitely. Ask what the person is agreeing to, say where it will appear, and note the answer. Where a clip may be reused in promotional material or a later piece, that is a separate conversation, and it is easier to have at capture than a year afterwards.
Children and vulnerable adults require a decision above reporter level about who may consent and what may be shown. Recordings of people in distress carry a further judgement about whether publication serves the reader or only dramatises the event. Files are personal data, so retention, storage and access follow the same discipline as notes: keep what you can justify keeping, securely, for a stated period.
08
8. File hygiene and pre-publication checks
Name files at capture with date, place and subject in a consistent pattern, because a folder of camera-generated numbers becomes unusable within a week. Keep originals separate from edited versions. Record where each published file came from, so that a later query about provenance has a documented answer rather than a recollection.
Before publishing, run a short pass over every asset: caption present and checked, alt text written, transcript or captions attached, credit and licence recorded, original preserved, and consent noted. Strip location metadata from images where publishing it would expose a person or a place that should not be identified. That last step is easy to forget and hard to undo.
Formats, minimum kit and the duty attached
| Format | Minimum kit | Text alternative | Chief risk |
|---|---|---|---|
| Still photograph | Phone camera, steady position | Caption plus alt text | Wrong or missing caption |
| Interview audio | Recorder plus clip microphone, room tone | Corrected transcript | Clipped level or hard room |
| Vertical video | Phone, support, external microphone | Synchronised captions with speaker labels | Inaudible speech and padded length |
| Chart or diagram | Data with definitions and source | Alt text carrying the finding | Alt text describing shape rather than result |
Checklist before publishing any media file
- Caption written from the scene, naming who, what, where and when.
- Alt text written to carry the information, not to repeat the caption.
- Transcript or captions attached and corrected by hand for names and figures.
- Original file preserved unaltered, with edits saved separately.
- Credit, licence or permission recorded for anything not created by you.
- Consent noted, including what the person was told about where it would appear.
- Location metadata reviewed and stripped where publishing it would expose anyone.
Questions readers ask
Is automatic transcription good enough to publish?
Not without correction. It is a useful first pass and it reliably mishears names, place names, figures and technical terms, which are exactly the words a reader will quote.
How much editing of a news photograph is acceptable?
Adjustments that represent the scene as it appeared are normal. Adding, removing, moving or duplicating content is not, and cropping that changes what the image shows is manipulation.
Do short clips really need captions?
Yes. Any published speech needs a text alternative for deaf readers, for readers in noisy places, and so the words can be found and quoted.