Accessibility with subtitles and transcripts: a practical guide

For years, subtitles were seen as an extra for foreign films. Today they’re a normal part of any video: they’re switched on by people who don’t hear well, people on public transport without headphones, language learners and anyone who simply wants to be sure they don’t miss a name. For many organisations they’re also an obligation: accessibility rules require captions on videos and a text alternative for audio. Doing them well isn’t hard, but there are rules. A subtitle that goes by too fast, splits a sentence down the middle or doesn’t say who’s speaking is almost as unhelpful as none at all. This guide focuses on the practical side: how to decide which type of subtitles you need, which rules to follow so they read well and how to check the result before publishing.

Who they’re for

• Deaf people: for them, subtitles are the only way to access what’s said. They also need to know who’s speaking and which sounds matter.
• Hard-of-hearing people, many of them older, who hear something but not enough to follow a fast conversation, accents or background noise.
• People who don’t speak the language well: reading and listening at the same time helps understanding, and a translated track opens the content to people who don’t speak it.
• People watching with the sound off: on a phone, in an open-plan office or on public transport. On social media, a large share of videos are played this way, and without subtitles the video simply tells them nothing.
• People with attention or processing difficulties, for whom the text reinforces what they hear and lets them reread.

Types of subtitles

• Dialogue subtitles: they only cover what’s said. They’re the usual choice for translating into another language, and assume the reader can hear everything else.
• Captions or SDH (subtitles for the deaf and hard of hearing): as well as dialogue, they show who’s speaking when it isn’t obvious and relevant sounds, such as music, laughter, a phone or applause.
• Closed: they’re in a separate file (SRT or VTT) and each viewer turns them on or off in the player. They can be corrected without exporting the video again and can be offered in several languages.
• Open or burnt-in: they’re part of the picture and always visible. Useful on platforms that don’t accept subtitle files, but they can’t be turned off or resized.
• Live: generated as people speak, at an event, a lesson or a public session. They arrive with a slight delay and with some errors there’s no time to fix.

A transcript is something else: the full text of what’s said, in a document you read without the video. It doesn’t replace captions on a video, but it’s the natural alternative for audio-only content such as a podcast or a meeting recording, and it’s very useful for searching within the content or quoting it. To choose the type of subtitles, think about where the video will be watched: on a website or platform that accepts files, closed subtitles are almost always the best option; on a platform that doesn’t, burnt-in ones; and if your audience includes deaf people, add speaker and sound information even if the video is in your own language.

What the standards ask for

The international reference is the Web Content Accessibility Guidelines (WCAG 2.x). Success criterion 1.2.2 requires captions for audio in prerecorded video (level A) and 1.2.4 requires captions for live audio (level AA). Other criteria in the same guideline require text alternatives for audio-only content. The guidelines don’t say how many characters a line should have or how fast subtitles should go; that’s set by industry style guides, which largely agree with each other. In the European Union, accessibility rules for public sector websites and apps are based on these guidelines, and more and more private sectors have similar obligations. In Spain, the UNE 153010 standard is the reference for subtitling for deaf and hard-of-hearing people: it covers how to show who’s speaking, how to represent sounds and how to present the text. If you work for television or for a client that requires it, it’s worth reading in full; for everything else, the rules in the following sections are a good basis.

Quality rules

• Two lines at most, of 42 characters each. More text forces people to stop watching the picture.
• Between 1 and 7 seconds per subtitle. Under a second leaves no time to read it; over seven makes people reread or think the video has frozen.
• Around 17 characters per second is a comfortable reading speed, and 21 is a reasonable maximum. If a subtitle goes over that, shorten the text or give it more time.
• Break at the end of a sentence wherever possible, and otherwise at a comma or a natural pause.
• Don’t leave articles or prepositions hanging at the end of a line, such as “the”, “of” or “and”: they’re harder to read.
• A new subtitle when the speaker changes, so two voices don’t share the same block.
• Sync: the subtitle appears when the voice starts and disappears shortly after it ends, not before.

To work out the speed, count the characters in the subtitle, including spaces, and divide by the seconds it’s on screen; most editors do it for you and warn you when it’s too fast. An example: a subtitle with two full lines has about 84 characters. At 17 characters per second it needs about 5 seconds on screen; at 21, about 4. If the person is speaking fast and there are only 3 seconds, it won’t all fit: you have to condense without changing the meaning or split the text into two subtitles. Captions for deaf and hard-of-hearing viewers tend to stay closer to the exact words, so condense with care.

Where to break

The most useful rule is that each subtitle should be a unit of meaning. “The budget was approved / at Tuesday’s meeting” reads well; “The budget was approved at / Tuesday’s meeting” makes people wait for the second line to understand the first. The same goes between subtitles: it’s better for a subtitle to end at a full stop or comma than in the middle of a phrase. When two lines aren’t the same length, a shorter top line, like a pyramid, usually reads better, although that’s a preference rather than a strict rule. Also avoid splitting a first name and surname, a number and its unit or a verb and its negative across lines. And when there’s a shot change in the video, ideally the subtitle shouldn’t run across it.

Who’s speaking and what can be heard

In captions for deaf and hard-of-hearing viewers, dialogue isn’t enough. If someone speaks off screen, or there are several people in a meeting and you can’t see who says what, it needs to be shown. The same goes for sounds that carry information: laughter that changes the meaning of a sentence, an alarm, music that sets the tone. In a recorded meeting or a panel discussion, where several people speak from the same shot, showing who’s speaking is almost always necessary. You don’t need to describe everything, only what someone who can’t hear needs to understand what’s happening. These are the most common conventions:

• The speaker’s name in brackets or capitals at the start of the subtitle, only when they can’t be seen or there may be doubt.
• A dash at the start of each line when two people speak in the same subtitle.
• Sounds in brackets in lower case, with a short description: (laughter), (phone rings), (applause).
• Music shown with its function or its title if it matters, not with a note every time it plays in the background.

The UNE 153010 standard sets out these conventions and a few more. Whichever convention you choose, keep it consistent throughout the video and across every video in the same series.

Transcripts for audio content

A podcast, a recorded interview or a training audio has no picture, so it doesn’t get subtitles: the alternative is a transcript. A good transcript shows who’s speaking in each paragraph, includes important sounds in the same way as captions and is corrected, especially names and numbers. Publish it on the same page as the audio, as text that can be selected and read with a screen reader, not as an image or only as a scanned PDF. Timestamps at regular intervals or at each change of topic help people who want to jump to a specific point. And if the audio also has a video version, as many podcasts do, the transcript complements the captions rather than replacing them. If the content is an internal meeting that isn’t published, the transcript also helps anyone who couldn’t attend or needs to go over something carefully.

Live events

A conference, a lesson or a public session streamed live is a different challenge: there’s no time to correct. Live captions can be produced by a person who types or respeaks (stenography or respeaking) or by automatic speech recognition. The first gives higher quality and costs more; the second is immediate and affordable, with more errors in names and noisy passages. If you use automatic captions, tell the audience so they know there may be errors. Either way, preparation makes the difference:

• Good sound: a microphone close to the speaker and no background speakers bleeding into the recording.
• Vocabulary in advance: speakers’ names, acronyms and technical terms, so the system or the person knows them.
• An easy-to-find link to the captions, announced at the start and on screen, for people following on their phones.
• After the event, publish the recording with reviewed captions and a corrected transcript.

Checklist

1. Every video with speech has captions, and every recorded audio has a transcript.
2. Subtitles have at most two lines of 42 characters.
3. No subtitle lasts less than 1 second or more than 7.
4. Reading speed never goes above 21 characters per second, and most subtitles stay close to 17.
5. Breaks fall at sentence ends, commas or pauses, and no line ends in “the”, “of” or “and”.
6. The speaker is shown when they can’t be seen or there may be doubt.
7. Relevant sounds are described in brackets.
8. Names, acronyms and numbers have been checked against the audio.
9. There are tracks in other languages if your audience needs them.
10. You’ve watched the whole video with captions on at least once before publishing.

Generate subtitles for free

If you have a video and want to start now, Echo’s free subtitle generator does it without an account: upload the video or audio, it transcribes it and gives you an SRT or VTT file following the rules in this guide, for the first 10 minutes of each file. Check the result before publishing, especially names and numbers. The steps for uploading them to each platform are in How to add subtitles to a video.

With Echo: editor, languages and live

For longer videos or regular publishing, Echo creates subtitles from the transcript, already split according to the rules above: two lines of 42 characters, between 1 and 7 seconds, 17 characters per second comfortable and 21 at most, with breaks at speaker changes, sentence ends, commas and pauses, and no lines ending in words like “the” or “of”.

• Subtitle editor: correct the text and timings, with undo and redo. It warns you if a subtitle reads too fast, and you can add speaker names or sounds by hand.
• Translated tracks: add several languages at once. The whole sentence is translated and spread across its subtitles, with the same timings as the original. Download each language as SRT or VTT, or all of them in a ZIP.
• Video with burnt-in subtitles, in the language you choose, for platforms that don’t accept files.
• Live translated captions for guests: while you transcribe with Live, you share a link and up to 300 people read what’s being said without an account, each in their own language and at the text size they prefer.
• Synthetic-voice dubbing in Spanish, English, French, Italian or Portuguese, with a voice for each speaker and the original quieter underneath or removed.
• Transcripts for your audio: download as Word, PDF or TXT, with each speaker’s name and timestamps if you like.

Transcription and translation run on Edorteam’s own servers in the European Union, and the AI isn’t trained on your videos. You can start on the Free plan: see pricing. If you work in a public authority that publishes its sessions, you may also find the guide on minutes in the public sector useful.