Subtitles and Captions: Craft, Conventions and Common Pitfalls

Subtitles and captions are often confused, yet they serve different audiences. Subtitles generally assume that the viewer can hear the audio but does not understand the language, so they translate the dialogue. Captions assume that the viewer may not be able to hear, so they transcribe speech and also describe relevant sounds, like music cues or doors slamming. Many platforms blur these terms, but the distinction is useful when planning.

Why they matter

Subtitles open content to people who speak other languages. Captions support viewers who are deaf or hard of hearing, and also those watching in noisy places or in situations where sound is off. Clear text also helps learners and viewers who find fast speech difficult to follow.

Basic craft principles

• Brevity: viewers must read while also watching the picture, so condense without losing the meaning. Subtitles are rarely a word-for-word transcript.
• Line length and breaks: keep to one or two short lines, and break at natural phrase boundaries rather than in the middle of a unit of meaning.
• Timing: each caption should appear when the speech begins and stay long enough to read, then disappear before it overstays.
• Synchronisation with cuts: try not to carry a caption across a shot change unless necessary, as it makes viewers re-read.
• Consistency: follow a style guide for punctuation, numbers, names and how to mark speakers.

Reading speed

Style guides specify maximum characters per second and per line, and different guides set different limits, particularly for children’s content. Because the best figure depends on audience and language, check the guidance of your platform or client rather than relying on a single number.

The workflow

1. Transcription: produce an accurate text of the dialogue, either manually or with automatic speech recognition followed by human correction.
2. Segmenting and timing: split the text into caption units and set in and out times.
3. Translation or adaptation: for subtitles in another language, translate with attention to length and rhythm.
4. Review: watch the video with the captions on, checking spelling, timing and readability.
5. Export: choose a suitable file format for the platform; common options include simple text-based caption files.

Automatic captions: helpful but imperfect

Speech recognition can save significant time, but errors are common with accents, background noise, overlapping speakers, names and technical terms. Always proofread automatically generated captions before publishing, especially for content where accuracy matters.

Dealing with tricky material

Songs and sound effects

Captions may indicate music or key sounds in brackets. Decide how much detail serves the story.

Names and cultural terms

Keep names consistent. For cultural references, you might keep the original term, adapt it, or add a brief gloss, depending on space and audience.

Swearing and tone

Register matters. Decide, with the client or editor, how strongly to render offensive language in each context.

Presentation and accessibility

• Use a clear sans-serif font with good contrast, such as light text over a subtle dark box.
• Keep captions clear of important on-screen graphics.
• Offer users the option to adjust size and colour where players allow.

Good captions go unnoticed because they let people follow the story effortlessly. That ease is the result of many small decisions about timing, wording and layout.

Watch without sound, then with

A useful final check is to view the finished video once with the sound off, relying only on the text, and once with sound while reading along. Gaps in meaning, awkward breaks and mistimed lines become much easier to notice from both perspectives.