How to add captions to a video

1. Captions, subtitles, open and closed: what the words mean

The words get used loosely, and the difference matters when a client or a platform asks for one specifically.

• Captions are the spoken words written on screen in the same language, usually with the sounds that matter too: [laughter], [phone rings], [music stops]. They are written for someone who can't hear the audio.
• Subtitles assume the viewer can hear the sound but doesn't understand the language, so they carry the dialogue, often translated, and leave out the sound cues.
• Subtitles for the deaf and hard of hearing (SDH) combine the two: dialogue plus sound cues. In British English, “subtitles” is often used for all of the above.
• Open captions are part of the picture. They are burned in and always visible.
• Closed captions are a separate track the viewer can switch on or off.

Most short company and social video uses same-language captions, so that is what this guide mostly calls them. Everything below applies to subtitles too.

2. Who captions are for

Captions started as an accessibility tool, and that is still the reason to do them properly rather than quickly. A viewer who is deaf or hard of hearing depends on them for everything the audio carries: who is speaking, what they said and what else can be heard.

• Caption the words that are actually said, not a tidied summary of them.
• Name the speaker when it isn't obvious who is talking, for example when they are off screen: “MAYA: We moved in March.”
• Describe sounds that change the meaning: a laugh after a line, an alarm, music that sets the mood. Skip background noise that doesn't.
• Keep captions on screen long enough to read, and in sync with the speech.

The same captions also serve people watching in a noisy place, in an open-plan office with the sound off, or in a language they read better than they hear. If your organisation has an accessibility policy or a legal requirement, check what it asks for; requirements differ by country and by sector.

3. Burned-in captions or an SRT file?

There are two ways to deliver captions, and you will often want both.

Burned-in (open) captions are rendered into the video. They look exactly how you designed them, on every device, with no setting to switch on. The cost: they can't be turned off, can't be corrected without a new render, and can't be swapped for another language.

A sidecar file sits beside the video and the player draws the captions. The two common formats are SRT and WebVTT. Both are plain text you can open in any text editor. An SRT cue looks like this: a number, a time range such as 00:00:04,200 --> 00:00:06,900, one or two lines of text, and a blank line. WebVTT is very similar: the file starts with the word WEBVTT and the times use a full stop instead of a comma before the milliseconds.

• Feeds where most people scroll with the sound off: burned-in captions are the safe choice, because you control the look and they show without the viewer doing anything.
• Video platforms and players that let viewers switch captions on: upload an SRT or WebVTT file. Viewers can turn it off, the platform can index the text, and you can fix a typo without re-uploading the video.
• Websites and learning platforms: usually a WebVTT or SRT file. Check which one your player accepts.
• Handing the edit to someone else: always send the SRT too, so the captions stay editable.

4. Start from an accurate transcript

Good captions start with good text. Most editing apps and plenty of standalone tools now transcribe speech automatically, and the result is a useful first draft. It is not a finished one.

• Transcribe the final edit, not the raw footage, so the captions follow the cut.
• Check every name, place, product name and job title against something written. These are where automatic transcripts go wrong most often.
• Check numbers. Decide whether you write “10 percent” or “10%” and stay consistent.
• Remove filler words such as “um” and “uh” unless they carry meaning, but keep the speaker's words and order otherwise.
• Listen with headphones to anything mumbled, overlapping or under music before you guess.

Read the captions with the sound on from start to finish at least once. It is the only reliable way to catch a word that was heard wrong but still looks like a real sentence.

5. Line length and reading speed

A caption the viewer can't finish reading is worse than none, because it pulls their eyes off the picture for nothing. Broadcasters and streaming services publish their own style guides, and these figures are common guidelines drawn from that tradition, not fixed rules:

• One or two lines on screen at a time, never three.
• Around 32 to 42 characters per line for a landscape video. Fewer for vertical video, where the frame is narrow.
• A reading speed in the region of 15 to 20 characters per second. If a cue needs more than that, shorten it or give it more time.
• At least about a second on screen for even the shortest cue.

Fast talkers make this hard. You can lightly condense a caption, dropping a repeated word or a false start, as long as the meaning and the speaker's voice stay intact. If you are captioning for accessibility, keep condensing to a minimum: a deaf viewer should get what a hearing viewer gets.

6. Timing and line breaks

Captions should appear when the words start and leave when they end. A cue that arrives late makes the viewer feel behind; one that arrives early gives away the punchline.

• Start each cue as the first word is spoken. End it shortly after the last word, but not so long that it overlaps the next line.
• Break cues at natural pauses: the end of a sentence, a comma, a breath.
• Inside a cue, break the line between phrases, not inside one. “We moved the whole team / to Tampere in March” reads better than “We moved the whole / team to Tampere in March”.
• Don't let a caption hang over a cut to a different speaker or scene if you can avoid it. Ending the cue at the cut reads more cleanly.
• When two people speak in one cue, start each person's line with a dash on its own line.

If you trim the start of a video after making the captions, every cue shifts. Re-export them from the edit, or move the whole file by the same amount: the free SRT shift tool moves every subtitle in an SRT file earlier or later in one go.

7. Caption style for social and everything else

Burned-in captions are part of the design of the video, so treat them like a title: readable first, on-brand second.

• Size: large enough to read on a phone held at arm's length. Preview it on a phone before you render.
• Contrast: white or light text with a dark outline, a shadow or a semi-transparent box behind it stays readable over any picture. Coloured text on a busy background doesn't.
• Font: a plain sans-serif. Decorative fonts slow reading down.
• Position: the lower part of the frame is the convention, but keep clear of the edges. Vertical feeds put buttons, names and descriptions over parts of the picture, so leave a safe margin and check the result in the app you post to.
• Lower thirds: never put captions on top of a name strap. Move the captions up while the lower third is on screen, or place the strap higher.
• Word-by-word highlight styles are popular on social feeds. Use them for short, energetic clips; for anything longer, plain full lines are easier to follow.

Keep one style for a whole video, and ideally across a series. The captions should look the same in your short social clips as they do in the longer film they came from.

8. Translating captions

Translated subtitles start from finished, checked captions in the original language, not from the raw transcript. Fix the source first, then translate.

• Keep the timing from the original file and translate the text cue by cue. The time codes in an SRT file don't need to change.
• Expect the text to grow or shrink. Some languages need noticeably more characters than English for the same sentence, so a translated line may need shortening to stay readable.
• Machine translation is a reasonable first draft. Have someone who speaks the language read it against the video, especially names, product terms and anything said as a joke.
• Deliver one file per language, and name it clearly, for example film.en.srt and film.fi.srt.

For a multilingual audience, separate subtitle files are usually better than burning several languages into one picture. Viewers pick their own, and you only render the video once.

9. Captions checklist before you publish

• Watch the whole video with the sound on and read every caption.
• Watch it again with the sound off. Can you follow it?
• Check every name, number and product term against something written.
• No cue is longer than two lines or too fast to read.
• Cues start with the speech and don't hang over cuts.
• Captions never sit on a lower third, a logo or a platform's buttons.
• Speaker names and meaningful sounds are captioned where the viewer needs them.
• The SRT or WebVTT file matches the final edit, not an earlier version.
• Translated files are checked by someone who speaks the language.

If the video is an interview, captions are one step of a longer job; how to edit an interview video covers the rest.