Record your whole screen, a single window, or one browser tab, with your microphone and the shared source's own audio if you want them. There is no extension to install, no account, and no upload — the recording is assembled in the page and stays on your machine until you download it, which is the point when you are capturing internal tooling, a client's dashboard or anything else you would not hand to a free web service. Microphone and system audio are mixed properly through a Web Audio graph rather than both being attached to the recorder, because a recorder given two audio tracks silently keeps only the first — the reason so many browser recordings come back missing the narration. You get a live timer, pause and resume, quality presets, and a preview before you commit to the download.
| Preset | Bitrate | Roughly per minute | Best for |
|---|---|---|---|
| High | 8 Mbps | ~60 MB | Text, code, fine detail |
| Balanced | 4 Mbps | ~30 MB | Most walkthroughs |
| Small | 1.5 Mbps | ~11 MB | Long recordings, sharing by email |
No. The stream is captured and encoded in the page, and the result lives as a temporary object URL in your browser until you download it or close the tab. Nothing is transmitted — which matters when what is on screen is internal tooling, customer data, or a client's system.
Usually, but it depends on what you share. Chrome and Edge offer to include audio when you share a tab, and sometimes a whole screen; Firefox's support is narrower and macOS restricts system audio heavily. When both are available they are mixed together properly, so you do not lose one of them.
Because MediaRecorder records only the first audio track it is given. Attaching both a system track and a microphone track looks correct and silently drops one. Mixing them into a single track through Web Audio first is the fix, and it is what happens here.
Chrome, Edge and Firefox on desktop all support screen capture. Safari's support is limited and most mobile browsers do not offer it at all — mobile operating systems provide their own screen recording instead. The page tells you up front if your browser cannot do it.
It is what browsers record natively, and VP9 compresses screen content well. Most editors and platforms accept WebM directly. If you need MP4 specifically, convert it afterwards — and on Safari, where WebM recording is unavailable, MP4 is used instead automatically.
As long as your memory allows. Data is collected in chunks and only assembled when you stop, which avoids holding the whole recording twice, but a very long capture at high quality still adds up. The Small preset is the sensible choice for anything lengthy.
Capture uses getDisplayMedia for the screen and, when a microphone is requested, getUserMedia alongside it. The two are separate streams, so their tracks are combined into one MediaStream before recording — and when the shared source also carries audio, that system track and the microphone track are mixed through a Web Audio graph rather than simply both being added, because MediaRecorder records only the first audio track it is given and the second would be silently dropped. The container is resolved before recording starts by asking MediaRecorder which types it supports, preferring VP9 for its noticeably better compression on screen content, then VP8, then MP4 for browsers that will not record WebM at all. Passing an unsupported type throws, and passing none lets the browser pick something it does not necessarily name in the resulting blob, so it has to be chosen explicitly. Data arrives as timesliced chunks and is only assembled into a Blob on stop, which keeps a long recording from being held twice in memory. A listener on the video track's ended event stops the recording when the user presses the browser's own stop-sharing button, which is otherwise a common way to end up with a file that never finalises. Nothing is uploaded — the recording exists only as a local object URL until it is downloaded.