Accessible Research Content: Captions, Transcripts and Alt Text
A practical checklist for making the media around your paper usable by more people: captions for video, transcripts for audio, alt text and long descriptions for figures, and colour contrast. Each step is tied to the W3C guidance it comes from.
You finish a video abstract, upload it, and the platform adds automatic captions. They look fine at a glance. Then you notice one line where a "not" has gone missing, and your careful null result now says the opposite.
Accessibility work for research media is mostly this kind of checking. It is not a separate project, and most of it uses text you already wrote for the paper. This guide walks through the four things that matter most for the media around one paper: captions, transcripts, alt text and contrast.
The Standard Behind This, in Two Minutes
The reference point is the W3C's Web Content Accessibility Guidelines (WCAG). WCAG 2.2 was published on 5 October 2023, with an update on 12 December 2024. It has 13 guidelines organized under four principles (perceivable, operable, understandable and robust), and each guideline has testable success criteria at three levels: A, AA and AAA.
WCAG 2.2 is also an ISO standard, ISO/IEC 40500:2025, and the W3C notes that content meeting WCAG 2.2 also meets WCAG 2.1 and 2.0. If your university or publisher has a web accessibility policy, check which WCAG version and level it expects before you start.
What Each Format Needs
The W3C's guide to making audio and video media accessible breaks media down into captions, transcripts, description of visual information, sign language and accessible media players. Mapped onto the formats researchers usually make for a paper, it looks like this (prerecorded media, which is what a paper's media almost always is):
| Format | What to provide | WCAG reference |
|---|---|---|
| Narrated video abstract | Captions | 1.2.2 Captions (Prerecorded), Level A |
| Narrated video abstract | Description of visual information, integrated into the narration or as audio description (a descriptive transcript alone covers Level A only) | 1.2.3 at Level A (description or transcript); 1.2.5 Audio Description at Level AA (description required) |
| Audio summary or podcast episode | Transcript | 1.2.1 Audio-only and Video-only, Level A |
| Silent animation or video | Descriptive transcript or an audio track of description | 1.2.1, Level A |
| Figures, graphical abstract, infographic | Text alternative; a long description for charts and diagrams | W3C images tutorial |
| Any text on slides or graphics | Contrast of at least 4.5:1, and colour never the only cue | 1.4.3 at Level AA; 1.4.1 at Level A |
The rest of this guide takes those rows one at a time.
Captions: Check Every Line
The W3C's captions guidance is blunt: automatically generated captions do not meet user needs or accessibility requirements unless they are confirmed to be fully accurate, and usually they need significant editing. Its own example turns "Broil on high for 4 to 5 minutes. You should not preheat the oven." into "Broil on high for 45 minutes. You should know to preheat the oven."
Research narration is full of the words that break this way: numbers, units, negations, drug names, gene names and acronyms. Treat the automatic file as a first draft and correct it against your script.
Two terms are worth keeping straight. The same page uses captions for the same language as the speech, and subtitles for speech translated into another language. Captions are needed for accessibility, while subtitles in other languages are not directly an accessibility accommodation, even though they widen who can follow the video.
Transcripts: You Probably Have One Already
For audio-only content, such as an audio summary of your paper or a podcast episode, the W3C's transcripts guidance places a transcript at WCAG Level A. A basic transcript is a text version of the speech and non-speech audio. A descriptive transcript also describes the visual information, and the W3C says descriptive transcripts are needed to provide video content to people who are both Deaf and blind.
If your audio or video was scripted, the script is most of the transcript. If you made captions first, most caption tools can export a plain text transcript, and the W3C notes you will usually need to add visual information, such as text shown on screen and who is speaking. Put the transcript in HTML on the same page, easy to find from the player.
Write the visuals into the script
The W3C's page on description of visual information describes integrated description: the speaker says what the viewer needs to know from the visuals as part of the main script. The W3C says that for some types of video, planning it in before recording means you do not need a separate description, so there is no additional cost. For a narrated video abstract, this is the simplest route to the description that Level AA asks for. For a video abstract, that means saying "the treated group recovered faster" rather than "as you can see here". The same page says all text in the video, including titles and a link shown at the end, should be available in the main audio, in the separate description, or clearly with the video.
Alt Text for Figures and Infographics
The W3C images tutorial sorts images by purpose. Informative images need at least a short description of the essential information. Decorative images get an empty alt attribute. Images of text should have a text alternative with the same words, and the tutorial advises avoiding text in images unless it is a logo.
Most figures from a paper are what the W3C calls complex images: graphs, charts, diagrams and maps. These need two parts: a short description that identifies the image and says where the long description is, and a long description that gives the essential information, such as scales, values, relationships and trends. The tutorial also suggests making long descriptions available to everyone, for example as part of the main content, and summarizing the figure's point in the surrounding text.
A worked example: one figure, two parts
Say your psychology paper has a line chart of reaction times across four testing sessions for two groups.
- Short alt text: "Line chart of mean reaction time across four sessions for the training and control groups. Full description below the figure."
- Long description, under the figure: one paragraph naming the axes and units, giving each group's values per session (copied from your results table, not re-estimated from the chart), and stating the trend in words.
- Surrounding text: one sentence that says what the figure shows, the way you would say it in a talk.
Everything in those three pieces already exists in your paper. The work is moving it next to the image.
Colour and Contrast
WCAG's contrast criterion (1.4.3, Level AA) asks for a contrast ratio of at least 4.5:1 for text and 3:1 for large text. The W3C treats 18 point text or 14 point bold text as large, and says the ratios should not be rounded: 4.499:1 does not meet 4.5:1.
The use of colour criterion (1.4.1, Level A) adds that colour must not be the only visual means of conveying information. On a graphical abstract or infographic, that means labelling groups directly or adding shape or pattern, not relying on a red line and a green line.
A Ten-Minute Pass Before You Post
- Correct the captions line by line against your script, paying attention to numbers and negations.
- Publish a transcript, built from the script or the caption file, on the same page as the audio or video.
- Check the script says aloud anything a viewer would otherwise only see.
- Give every figure and infographic alt text, and a long description for charts.
- Check text contrast on slides and graphics, and make sure no message relies on colour alone.
Where Amplify Fits
Amplify does not change what WCAG asks of you, and nothing in this checklist goes away. What it changes is the starting point. From the PDF you upload, and nothing else, it makes a video abstract, a vertical short, an audio brief and a one-page infographic, each carrying the title, the authors and a link back to the paper of record.
Each format arrives as three variants that you can edit line by line, so you can tighten the wording before anything goes public, and nothing publishes until you approve it. The embeddable widget places the summaries next to the audio brief, the video and the infographic on your own page, so a reader who cannot listen to or watch a format still has the paper's findings in text. For more on reaching readers in their own language, see multilingual research summaries, and for the audio side, choosing an audio format for your paper.
Accessible media is clearer media: a corrected caption, a script that says what the chart shows and a figure described in words help every reader, not only the ones who need them. Build the checks into the same afternoon you publish.
Quick answers
Are automatic captions good enough for a video abstract?
Not on their own. The W3C says automatically generated captions do not meet user needs or accessibility requirements unless they are confirmed to be fully accurate, and usually need significant editing. Missing a single word such as "not" can make a caption contradict what was said, so read every line against your script before publishing.
Does a research podcast or audio summary need a transcript?
Yes. Under WCAG, prerecorded audio-only content needs a text alternative that presents equivalent information, which is a transcript, at Level A. If you scripted the audio, you already have most of it: publish the script as text beside the player and add speaker names.
How do I write alt text for a chart or figure from my paper?
Use two parts. The W3C calls charts and diagrams complex images, which need a short description that identifies the image and a long description that gives the essential information, such as the values and trends. It is also good practice to summarize the point of the figure in the surrounding text.
What colour contrast should text on a research graphic have?
At least 4.5:1 for normal text and 3:1 for large text, under WCAG 1.4.3 at Level AA. The W3C treats 18 point text, or 14 point bold, as large, and the ratios are not rounded, so 4.499:1 fails. Colour should also never be the only way information is shown.
