Video Transcripts: Write Down the Step the Camera Shows

A maker films a short tutorial on folding a paper gift box. Halfway through, she says, ‘Slide this underneath, then press here.’ Her hands show which flap goes under the other and where to press. The transcript records every word. A reader follows it to that point and has to stop: which flap, underneath what, and where should the pressure go? The words are accurate. The instruction is incomplete.

If you publish tutorials, demos or lessons, take a pass through the transcript for the things the camera explains on its own. The job is to make the necessary information available in text. You can start with existing captions and keep much of the original speech. The difference comes from noticing where the speaker points, pauses or relies on something visible to finish the sentence.

Separate the words from the information

The W3C Web Accessibility Initiative’s guide to transcripts, first published in September 2019 and updated in September 2024, distinguishes a basic transcript of relevant audio from a descriptive transcript that also includes visual information needed to understand the content. That distinction is useful even before you decide how to lay out the page. A transcript can be an accurate record of the sound and still leave a reader without the material they need to follow the lesson.

Look for moments when the picture does a job the words leave open. In a craft video, the picture may show the direction of a fold. In a screen demonstration, it may show which menu is open or the exact wording of a status message. In a lesson, it may show an equation the presenter points at without reading it. Write down the missing instruction or evidence at the point where it matters. A paragraph describing the studio décor will rarely help someone find the right flap.

For the paper box, a useful note might read: ‘Tuck the narrow side flap beneath the wide flap. Press along the fold where the two meet.’ That describes the operation. If the presenter only gestures and the angle makes the join hard to see, check another shot or ask the presenter what they did. A precise description based on a guess can be harder for a reader to detect than an obvious gap.

Make the added detail easy to find

Keep descriptions beside the spoken instruction they clarify. If the speaker says ‘Now compare the two’, put the comparison there, while the reader still knows which two things are in play. If an important label appears earlier on screen, introduce it before it is used. A reader should not have to open a second document or hunt for a note at the bottom to complete a step.

Use a consistent convention that distinguishes spoken words from editorial descriptions. For a short tutorial, separate paragraphs introduced by ‘On screen’ can work. For a longer lesson, descriptive headings can help someone return to the stage they need. Timestamps can be useful when readers also want to jump to the video, but a timestamp by itself gives them a location and leaves the instruction unstated. Keep any timings tied to the final cut.

Our guide to choosing a video poster frame looks at what an image tells people before they press play. A descriptive transcript solves a different problem: can a reader recover the information the picture carries after the lesson starts? The same frame can be useful in both jobs, but the transcript has to put its useful detail into words.

Try following the finished text

Read the transcript without playing the video. Better still, give it to someone who has never watched the clip. Ask them to identify the object or control at each step, what changes and how they would tell whether it worked. In the paper box example, ‘the narrow flap stays tucked when you release it’ gives a reader a check they can use at the workbench. Include that cue if the video demonstrates it and it matters to the result.

• Check references such as ‘this’, ‘there’ and ‘the one on the left’. Replace or explain them where the intended object would be unclear from text alone.
• Read any numbers, warnings or labels that appear only on screen. Include the ones that change how the viewer carries out the task or interprets the result.
• Open the published page and find the transcript from the video. Check the version matches the final edit and that the text is easy to read on a phone.

If the transcript begins as an AI draft, give the reviewer the final video as well as the text. Mark uncertain descriptions for review, especially where two objects look alike or a screen label is too small to read. Review the step that could send someone down the wrong path first. A transcript can retain a speaker’s style and still be precise about an action the speaker left to the camera.

The final check is simple: give the reader the text and see where they have to stop. Fix the missing action, label or result at that point. The transcript earns its place when someone can follow the lesson at their own pace, using the words on the page.

Source

• W3C Web Accessibility Initiative, Transcripts, updated 17 September 2024.