Captions are text of the spoken words and meaningful sounds, synchronized with the video. They serve people who are Deaf or hard of hearing, people in noisy or quiet places, and people processing a second language. Automatic captions are a draft, not a finished product; they miss names, acronyms, program terms and anything said quickly, and they rarely punctuate. Edit them, and include speaker changes and sounds that matter, such as “[phone rings].” Transcripts are a complete text version of the video that can be read without playing it. For a screen-reader user, a transcript is often faster than the video; for everyone, it is searchable and skimmable. A transcript for a demonstration video includes what happens on screen, not only what is said. Audio description is narration of visual content that matters and is not spoken aloud: what is on the slide, where the cursor goes, who entered the room. The simplest form of audio description is a narrator who says what they are doing, which is why re-recording “click here” as “select Save, at the bottom right of the assessment” often removes the need for a separate description track.
Slides fail in predictable ways: text typed into free-floating boxes instead of the layout placeholders, so a screen reader has no title and no order; images with no alt text; text pasted in as a picture; tiny type and low contrast because the slide was built for the presenter’s screen. Build on the built-in layouts, give every slide one unique title, check the reading order in the selection or outline pane, keep text as text with alt text on images, and use the accessibility checker your slide software provides as a first pass, not a final word. Send the deck ahead of the meeting as an accessible file, because slides on a screen are not readable by everyone in the room.
Forms are where people are most often stopped. Every field needs a label that is programmatically attached, not just visually nearby, so a screen reader announces “Date of birth, edit” rather than “edit.” Instructions and format examples come before the field, not after. Required fields are marked in text. Error messages appear next to the field, say what went wrong and what to do, and move focus to the problem. Grouped controls, such as a set of choices, share a group label. There is no time limit without a warning and a way to extend, and progress can be saved. A PDF form has all of this only if it was built as a tagged, fillable form with labeled fields; a scanned or flattened PDF is an image and needs a web form or an accessible alternative beside it.
