Live Captions

This page deals with live media, which includes both audio and video. The criterion states:

Captions are provided for all live audio content in synchronized media.

Since I first wrote a blog post about live captions, back in 2021, technology has moved on tremendously. I then wrote this page in 2024/25, but amended it in 2026, to reflect my current thinking on the subject.

Video Calls

If we go back to pre-pandemic times, video calling was usually between two or three people and we used platforms like Skype. Then Covid hit us. We were all locked down, working from home and the need for online communication tools suddenly changed completely. We began to use Zoom and Teams to meet with family, host quizzes and generally cope with the isolation of staying at home all the time.

Now, a new normal has emerged. The use of video calls has now become commonplace. Training is now more often a webinar, rather than a face-to-face event. The need for live captions in these platforms is much greater than it once was.

If you read the Understanding document for 1.2.4 Captions - Live, you will read that this criterion is not intended to cover two-way calls involving two or more participants. In such situations, the responsibility for providing captions is with the caller or the host... so the person who sends the Teams invite... rather than with the application itself. It does mention live training and webinars though.

Most of the video calling platforms now have some kind of live caption tool built in. This can usually be toggled on and off in one of the menus. Regular users are likely to have it set up to meet their needs.

Accuracy

Automated live captions are rarely completely accurate. They depend on the clarity of speech, background noise and interference, and even the tone and pitch of the speaker. If the speaker has a strong accent, this will usually affect the accuracy of captions.

This is where I think presenters can do more to enable the automated captions to be as accurate as possible. It is more difficult in a meeting with lots of people, where everyone is likely to talk and what they say isn't planned. However, if you are delivering a presentation or training, there are some things you can do to help.

• Plan what you are going to say. Even use a script, if possible. This reduces ums and ahs, and helps sentence punctuation in the live captions.
• Speak slower than you find comfortable. We all tend to speak more quickly than we realise. Slowing down really helps with the captions. Pause between sentences.
• Ask the audience to stay on mute unless they are speaking. Background sounds can interfere with the captioning.
• Invite anyone who is using captions to interrupt and let you know if they need you to speak more clearly or slow down.

Broadcasted Media

This success criterion was originally intended to address broadcasted media, such as live television.

The kind of live captions that are needed do not rely on automated tools, which are often inaccurate. Best practice is to use Communication Access Realtime Translation (CART). CART captions are produced by a trained captioner, using special equipment that allows them to produce captions quickly. It is still not perfect, as they are working at the speed of the speaker, but human-produced captions are far more accurate, as they can interpret the speech in ways that automated captions cannot.