Most real-time video projects get their accessibility review at the very end, usually after a user writes in to say they cannot find the mute button with a keyboard. By then the call screen is a tangle of absolutely positioned overlays, hover-only controls and status changes that only exist as colour. Fixing it late is slow and expensive. Building it in from the first sprint is mostly a matter of knowing where the traps are.
This checklist collects the issues that come up again and again in audits of live video interfaces. It is framework-agnostic and assumes you are aiming for WCAG 2.2 level AA, which is the bar most accessibility policies and procurement requirements point to. Each item explains what to check, why it matters and how to test it in a few minutes.
Controls and keyboard
1. Every call control is a real button
Mute, camera, screen share, leave and settings should be native button elements, not styled divs with click handlers. Native buttons get focus, Enter and Space activation and a role for assistive technology for free. When a control toggles state, expose it with aria-pressed so a screen reader announces "Mute, toggle button, pressed" rather than just "Mute".
2. Controls never disappear on a timer without a keyboard path
Auto-hiding toolbars look clean in a demo, but if the bar only reappears on mouse movement, keyboard and switch users are stranded. Show the toolbar whenever anything inside it receives focus, and keep it visible while focus stays there. Pressing Tab from anywhere on the call screen should land on the first control.
3. Shortcuts are documented and remappable
Single-key shortcuts such as M for mute are convenient, but WCAG 2.1.4 requires that character-key shortcuts can be turned off or remapped, or only work when a component has focus. Speech-input users trigger them by accident constantly. Offer a shortcuts panel in settings with an off switch, and prefer modifier combinations for destructive actions such as leaving the call.
4. Focus is visible and never hidden behind video
Video tiles, picture-in-picture windows and floating reaction bars love to sit on top of focused elements. WCAG 2.2 added 2.4.11, Focus Not Obscured, for exactly this reason. Test by tabbing through the whole screen with a self-view tile open and a participant pinned, and check that the focus ring is never covered.
Captions, audio and announcements
5. Live captions are a first-class feature
Captions for live audio are a level AA requirement under 1.2.4, and they help far more people than the deaf and hard-of-hearing users they are designed for: people in noisy places, non-native speakers and anyone on a poor connection. Whether you use an in-browser speech recognition API or a server-side transcription service, put the toggle on the main toolbar rather than three menus deep, and let users resize the caption text and move the caption area.
6. Call state changes are announced
When someone joins, leaves, raises a hand or starts sharing a screen, sighted users see it. Screen reader users need a polite aria-live region that announces short, plain messages such as "Sam joined" or "You are muted". Throttle these: in a large group a stream of join messages becomes noise, so batch them after a second or two.
7. Audio cues have a visual twin, and the reverse
A chime when someone knocks or joins is helpful for blind users and invisible to deaf users. A coloured speaking indicator around a tile is the opposite. Pair every audio cue with a visual one and every visual-only signal with text or an announcement, and let users switch the sounds off independently.
Visual design and motion
The visual layer of a call screen is where most quick wins live, because the problems are easy to see once you look for them.
| Check | WCAG 2.2 reference | Quick test |
|---|---|---|
| Text over video meets 4.5:1 contrast | 1.4.3 Contrast (Minimum) | Put a white wall and a bright window behind the camera and read the name labels |
| Icons and toggle states meet 3:1 | 1.4.11 Non-text Contrast | Compare muted and unmuted icons in greyscale |
| State is not shown by colour alone | 1.4.1 Use of Color | Turn on a greyscale filter and identify who is muted |
| Targets are at least 24 by 24 CSS pixels | 2.5.8 Target Size (Minimum) | Use the mobile layout and tap every control with a thumb |
| Layout works at 400% zoom | 1.4.10 Reflow | Zoom the browser and check that controls stay reachable |
8. Labels on tiles survive any background
Participant names sit on top of live video, which means the background changes every frame. A semi-opaque dark plate behind the label solves contrast for almost every scene; a text shadow alone does not. Truncate long names with a full name available on focus or hover, not just a tooltip that only mouse users can reach.
9. Motion respects user preferences
Animated reactions, confetti and tiles that slide around when the active speaker changes can trigger discomfort for people with vestibular disorders. Honour the prefers-reduced-motion media query: swap animations for simple fades or instant changes, and let users pin a speaker so the layout stops jumping.
Permissions, errors and onboarding
10. Device prompts are explained before they appear
Browser permission dialogs for camera and microphone are terse and appear without context. A short pre-permission screen that says what will be requested and why reduces denials for everyone, and it gives screen reader users a moment to orient before the browser takes focus. Whether the product is a team meeting tool or a social video chat app, this step is the same.
11. Errors say what happened and what to do
"Something went wrong" is not an error message. If the camera is in use by another app, the microphone permission was denied, or the network dropped, say so in text, move focus to the message or announce it, and offer a single clear next step. Keep the user in the call where possible rather than throwing them back to a lobby.
12. Matching and lobby flows are operable without a mouse
Products built around chance encounters, such as a random video chat with interest filters, often rely on swipe gestures and "next" buttons that fire on hover or drag. Every gesture needs a single-pointer or keyboard alternative under 2.5.1 and 2.5.7, and safety actions such as report and block must be reachable in two key presses or fewer, because they are used under stress.
How to fit this into a sprint
You do not need a dedicated accessibility team to make progress. A realistic plan for a small frontend team looks like this:
- Week one: run an automated scan on the call screen and fix the obvious items: missing labels, contrast, non-button controls.
- Week two: do a keyboard-only pass through a full call, from lobby to leaving, and log every place you get stuck.
- Week three: test with a screen reader on desktop and on mobile, focusing on announcements and focus order.
- Ongoing: add a keyboard and screen reader smoke test to the definition of done for any feature that touches the call screen.
Mistakes that keep showing up in audits
- Relying on automated scanners alone. They catch roughly the mechanical issues and miss most focus, announcement and flow problems.
- Hiding the leave button inside a menu to prevent accidental exits. Use a confirmation step instead.
- Announcing every speaker change through aria-live, which makes a screen reader talk over the actual conversation.
- Treating captions as a premium feature or an experiment flag that ships off by default.
Accessible real-time video is not a special discipline; it is ordinary frontend craft applied to an interface that changes every second. Start with real buttons, visible focus and honest announcements, add captions early, and test with a keyboard before every release. Your call screen will be easier to use for everyone, not just the people the guidelines were written for.
Further Reading
Discover more articles on similar topics across our network

Comments
Loading comments…