What the browser can capture
Screen recording in a browser uses the Screen Capture API, which requires an explicit user gesture and shows a picker offering the whole screen, a specific window, or a single browser tab. This permission model is deliberate: a page cannot start recording silently, and the browser displays an indicator while capture is active.
The three scopes differ in ways worth knowing. Capturing a tab records only that tab's content and follows it, so notifications and other windows never intrude — the safest choice for a demonstration. Capturing a window records that application only, and capturing the entire screen records everything, including anything that appears unexpectedly.
Audio capture is more constrained. System audio can generally be captured when sharing a whole screen or a Chrome tab, but not when sharing a window, and support varies by browser and operating system — on macOS, system audio capture is more limited than on Windows. Microphone audio is a separate stream and needs its own permission.
Settings that determine whether text is readable
Screen content is unlike camera footage: it is mostly static, with sharp text edges and flat colour. That means frame rate matters less and resolution matters more. Recording at 30 frames per second is ample for most demonstrations, and 60 is worth it only for genuinely smooth motion such as animation or gameplay.
Resolution should match the display rather than be scaled. Recording a high-density screen and exporting at half size makes small text unreadable, and enlarging afterwards makes it worse. If the recording will be viewed in a small embedded player, the answer is to increase the interface's font size or zoom level before recording, not to enlarge afterwards.
Bitrate governs whether text stays crisp. Video codecs are tuned for photographic content and discard exactly the high-frequency detail that text is made of, producing the shimmer around characters familiar from poor screen recordings. A generous bitrate, and where available a codec setting favouring detail over motion, addresses it.
Recording without leaking anything
Screen recordings disclose far more than intended with striking regularity. Notification banners arriving mid-recording have exposed private messages, browser tabs have revealed unrelated work, bookmark bars have shown internal tool names, and autocomplete dropdowns have displayed previously entered addresses and search terms.
Before recording: enable do-not-disturb, close unrelated tabs and applications, hide the bookmark bar, sign out of or switch to a clean profile, and clear anything on screen you would not publish. Recording a single tab rather than the whole screen eliminates most of these at once.
Watch the content itself too. API keys in a terminal, tokens in a URL, email addresses in a test account, and real customer data in a demonstration environment are all common leaks — and unlike a screenshot, a recording exposes everything that appears at any moment, including things visible only briefly. Review the finished recording before publishing, since that is the only way to catch what appeared for half a second.