How a Caption Gets Written, Checked and Fitted to a Phone Screen
A look at the working routine behind readable captions: transcribing, splitting lines, timing them to speech and checking the result on a small screen before export.
Priya Nair · · 7 minutes to read
Starting from the spoken words, not the screen
Every caption job in the course begins the same way: with the audio, played back slowly, and a plain text document open beside the timeline. Before any text appears over a picture, I transcribe what the speaker actually said, including the false starts, because those decisions about what to keep come later and should be made on purpose rather than by accident.
Event recordings are rarely clean. A hall echoes, a microphone drifts, someone coughs over a key phrase. When I cannot make out a word after three listens, I mark it and move on, then return with fresh ears. Guessing at a word and putting it on screen is the quickest way to misquote a speaker, and a misquote in a recap is hard to take back once it has been shared.
This is the first habit learners practise on the labelled sample projects: a transcript that is honest about its gaps. A caption can only be as accurate as the notes behind it, so we treat the transcript as the source of truth and the on-screen text as a careful selection from it.
- Play each clip at least twice before typing
- Mark unclear words instead of guessing
- Keep the raw transcript separate from the caption draft
Cutting lines to fit a small screen
Once the transcript is settled, I shorten. A sentence that reads well on paper often runs too long for a phone held at arm's length. The usual aim is two lines at most, with each line short enough to take in at a glance, and a break placed where a speaker would naturally pause rather than in the middle of a phrase.
Shortening is where meaning is most at risk. Dropping filler words is safe. Dropping a qualifier such as 'we hope to' or 'in some cases' is not, because it turns a modest remark into a promise. I read the trimmed line aloud against the original and ask whether the speaker would recognise it as theirs. If the answer is hesitant, the line goes back to being longer and gets split in two.
Learners often expect this to be a technical task. In practice it is closer to editing a letter: you remove what is spare and protect what carries the sense, and you accept that some lines simply need more time on screen.
- Aim for two lines, broken at natural pauses
- Keep qualifiers that change the meaning
- Read the shortened line against the original aloud
Timing captions to the voice
Timing is the part viewers feel without noticing. A caption that arrives a beat early gives away the point before the speaker makes it; one that lingers after the voice has moved on makes the next sentence feel unsupported. I place each caption so it appears just as the first word is spoken and clears shortly after the last.
Reading speed matters as much as sync. A short caption can leave quickly, but a longer one needs a moment to be read, even if the speaker finishes sooner. When the two conflict, I look at the edit rather than the text: sometimes a half-second of held picture solves it, and sometimes the line is simply too long and should be cut again.
On the course, this is practised with a stopwatch habit rather than a rule. Learners watch their own sequence with the sound off and see whether they can follow it. If a caption is gone before they finish reading, it is gone too soon for anyone else.
- Appear with the first word, clear soon after the last
- Allow reading time for longer lines
- Test the sequence with the sound off
Titles, names and lower thirds
Captions carry speech, but titles carry context: who is talking, where, and what the moment is about. I keep these plain. A name and a role in a clean, readable type, placed clear of faces and of the lower edge where some apps overlay their own buttons, is usually all a recap needs.
Spelling a name correctly is not a detail. I confirm names and job titles against the event's own programme or with the organiser before anything is exported, since on-screen text feels official in a way that spoken words do not. If a title cannot be confirmed, we leave it out and let the speaker introduce themselves in the clip.
Style stays consistent across a video. One typeface, one size for captions, one for titles, one colour that holds up against both light and dark footage. Learners are encouraged to resist decoration; a caption that draws attention to itself is competing with the person it is meant to serve.
- Confirm names and roles with the organiser
- Keep text clear of faces and platform buttons
- Use one typeface and a small set of sizes
Checking on the device people will actually use
Before export, I look at the video on a phone, not only on the editing monitor. Text that looks generous on a large screen can shrink into a smudge in a vertical frame, and a caption placed safely in a horizontal version may sit under a platform's controls in a square or vertical one.
This check is the reason the caption stage sits next to delivery in the course. Priya's sessions on formats and captions overlap for exactly this reason: each export size may need the same line moved, resized or re-broken. It is slower than reusing one version everywhere, but it avoids the common problem of a recap that works on one screen and fails on the rest.
I also play the final video once through with fresh attention and a printed transcript, ticking off each line. It takes a few minutes and has caught more slips than any software check, usually a swapped word or a stray character left over from an earlier draft.
- Preview on a real phone before exporting
- Re-check caption position in each frame shape
- Proofread the finished video against the transcript
What learners take from this routine
By the end of the captions module, learners have a short, repeatable process: transcribe, trim with care, time to the voice, confirm names, and check on a small screen. None of it needs special software, and all of it can be applied to the next event they are asked to cover.
The feedback rounds then test the habit on their own edit. Reviewers read captions the way an audience would, quickly and without context, and point out lines that are too long, too early or quietly changed from what was said. Most revisions are small, and that is the point: careful captions are the result of many small decisions made consistently.
The aim is not decoration but trust. A viewer who can follow the words easily, and who is never misled about what was said, is more likely to watch to the end and to believe what the recap shows.
Good captions are mostly patient work done before anyone sees the video: listening twice, cutting with care, timing to the voice and checking on a phone. None of it is glamorous, yet it decides whether a recap feels clear or careless. If you practise the routine on a short clip this autumn, you will find the habits carry over to every event you edit afterwards.