How to Screen Record a Video Playing on Your Screen

Recording video playback stutters, balloons file size, and sometimes goes black. Here is why, and how to capture on-screen video that stays watchable.

Most screen recordings capture a static interface: menus, forms, dashboards, a cursor moving between them. Then one day you need to record something that is already moving — a video player in your own product, a clip embedded in a slide deck, a motion prototype, a recorded Zoom call you are annotating. You hit record, play the video, stop, and watch it back. The interface around it looks fine. The video inside it looks terrible: uneven, jerky, occasionally stuttering in a rhythm that was definitely not there on screen.

Nothing is broken. Recording video playback is a fundamentally harder problem than recording a UI, and it fails in three predictable ways: motion, size, and sound. Here is what is actually happening and what to do about each.

Why On-Screen Video Stutters: Two Clocks That Don’t Divide

A video file has a frame rate baked into it. Film and most cinematic content is 24 fps. European broadcast content is 25 fps. Web video and most screen recordings are 30 fps. Game capture and sports are often 60 fps.

Recorded captures screen frames at 30 fps. That number is the capture clock, and it does not adapt to what is playing on your screen.

When those two numbers divide cleanly, you get a clean result. A 30 fps video sampled at 30 fps lands one captured frame per source frame — effectively a perfect copy. A 60 fps video sampled at 30 fps keeps every second frame, which is a well-behaved halving; motion is less fluid than the original but perfectly even.

When they don’t divide cleanly, you get judder. A 24 fps video sampled at 30 fps means the capture clock needs 30 samples from 24 available frames, so six of those frames get sampled twice and eighteen get sampled once. The result is a repeating pattern of held and advanced frames — the same cadence problem that makes film look slightly wrong on a 60Hz TV, except more visible because 30 is a smaller number to spread across. A 25 fps source is worse still: the pattern that repeats doesn’t line up with anything, and smooth pans turn into a visible shuffle.

This is why “my recording of a video is choppy” has nothing to do with your hardware most of the time. You can have a brand new machine, no dropped frames, a completely idle CPU, and still get juddery playback in the file, because resampling 24 into 30 cannot be done without repeating frames.

Two things follow from this:

You cannot fix it in the editor. Once a frame has been held for two sample intervals, the information about what happened in between is gone. Export renders at 60 fps, which gives synthesized motion — zooms, the redrawn cursor, text transitions — plenty of frames to work with, but it cannot reconstruct source frames the capture never saw. For more on how capture rate and export rate relate, see choosing the right frame rate.

You should decide whether the motion is the point. Most of the time it isn’t, and that opens up much better options than fighting the clock.

Record the Interface, Not the Playback

The single most effective technique for recording on-screen video: don’t record the playback at full speed if the playback isn’t your subject.

Ask what the viewer actually needs to see. In most cases where a video appears inside a screen recording, the video is context, not content. If you are demonstrating a video player, the point is the controls, the timeline scrubbing, the quality selector — not whether the clip inside plays fluidly. If you are walking through a motion prototype, the point is usually the design decision, not the exact easing curve.

So:

Scrub instead of play. Dragging the playhead through a video shows the content, demonstrates the player, and produces motion you control completely. Scrubbing is a cursor drag with a changing image behind it, which 30 fps capture handles well.

Pause on the frames that matter. Play for a second, pause, talk about what is on screen, advance. This reads as deliberate narration rather than a broken recording, and every paused frame is pixel-perfect.

Let the viewer watch the real video elsewhere. If the clip itself genuinely matters — a customer testimonial, a trailer, a conference talk — a recording of it playing inside a browser window is the worst possible delivery. Link to it. Your screen recording should cover everything around it.

Use a zoom to carry the moment instead. A zoom push into the player region is synthesized at export time and renders smoothly regardless of what the source footage does. In a case where you need to signal “watch this part,” a zoom plus a text overlay does more than 90 juddery frames of playback.

When you do need the playback in the recording, keep the clip short. Three to five seconds of imperfect motion inside a two-minute walkthrough reads as “there was a video there.” Forty seconds of it reads as a quality problem.

The Audio Side: System Sound, Separately

A video playing on screen has a soundtrack, and that soundtrack needs to come from system audio capture, not your microphone pointed at your speakers.

Turn on the System Audio toggle in the media controls before you start recording. This captures the sound your machine is playing at the source — clean, with no room reverb, no keyboard clatter, and no risk of feedback. On macOS this runs through ScreenCaptureKit with no extra audio driver to install; on Windows it uses the system loopback. Platform specifics are in the guides for recording with internal audio on Mac and recording with audio on Windows 11.

The important part is what happens afterward. Microphone and system audio are kept as separate tracks, each with its own volume in the editor, and each supporting volume segments you add by double-clicking the track. Volume runs from 0% to 200%, and either track can be muted outright.

That separation is what makes recording video playback workable:

  • Duck the video under your narration. Drop system audio to 20–30% for the stretch where you are talking over the clip, then bring it back up when you stop. This is the single most professional-sounding thing you can do with a recording that contains other people’s audio.
  • Mute the clip entirely where it adds nothing. A UI demo that happens to have a video playing in it does not need that video’s soundtrack competing with your voice.
  • Keep it loud where the clip is the point. If you are reacting to something, let it be audible and stop talking.

One caution: a recording that contains someone else’s audio is a recording that contains someone else’s audio. Music, trailers, and film clips carry the same licensing questions in a screen recording that they do anywhere else. Muting the system track is often the right answer for anything you plan to publish.

File Size Goes Up Fast, and That’s Expected

Screen recordings compress extremely well because most of the frame doesn’t change. A sidebar that sits still for forty seconds costs almost nothing to encode. That is why a ten-minute UI walkthrough can be a surprisingly small file.

Video playback destroys that advantage. Every pixel in the player region changes every frame, so the encoder has to describe genuinely new content constantly. A recording that is mostly video playback can be several times the size of an equivalent-length UI recording. Add a high-detail clip — confetti, water, film grain, a crowd — and it goes higher.

What to do about it at export time:

Choose the resolution deliberately. MP4 export offers 720p, 1080p, 1440p, 4K, Original, and Retina. Exporting a recording of a 1080p video playing in a window at 4K cannot add detail that the source video never had — you are spending bits on upscaled pixels. Match the export to the real resolution of your content.

Don’t reach for GIF. GIF has no interframe compression worth the name and a 256-colour palette, which is the worst possible combination for full-motion video. A few seconds of playback as a GIF will be enormous and look banded. GIF export caps at 480p or 720p with frame rates of 10, 15, or 24 fps for exactly this reason — it is a format for short, simple, high-contrast interactions. The GIF guide covers where it does work.

Trim ruthlessly before you export. The cheapest compression is deleting footage. If you recorded forty seconds of playback and need five, cut the other thirty-five before worrying about bitrate.

The general techniques in reducing screen recording file size all still apply — they just matter more here.

When the Video Records as a Black Rectangle

Sometimes the problem isn’t quality at all. You record a streaming video and get a recording where the interface is intact and the player is a black box.

That is DRM, and it is working correctly. Streaming services mark their playback surfaces as protected content, and both macOS and Windows capture APIs honour that marking at the operating system level — ScreenCaptureKit on macOS and Windows Graphics Capture on Windows. No setting in any recording tool changes this, because the exclusion happens before any recorder sees the frame. The same mechanism applies to some banking and enterprise apps that deliberately mark their windows as unrecordable.

If your own video is coming out black and no DRM is involved, it is usually a different issue — a permission that was never granted, or a GPU mismatch on a Windows laptop with switchable graphics. The troubleshooting guide walks through both.

Performance: Three Jobs at Once

Recording video playback asks your machine to decode a video, composite it to screen, capture that screen, and encode the capture — simultaneously. On a modern machine with hardware decode and hardware encode, that is usually fine. On a thermally constrained laptop running a browser with thirty tabs, it is where dropped frames actually start showing up.

The distinction matters when you are diagnosing. Judder from frame-rate mismatch is even — a repeating pattern that looks the same throughout the clip. Judder from a machine that can’t keep up is uneven — fine for a while, then a lurch, then fine again, often with audio drifting out of sync. If you are seeing the second kind, close what you don’t need, record a shorter region rather than a full 4K display, and work through the performance optimization guide.

One more practical note: play the video from a local file rather than streaming it when you have the choice. A buffering stall mid-recording is indistinguishable from a performance problem on playback, and you cannot edit it out without cutting the clip.

A Short Checklist

Before recording anything with video playing in it:

  1. Decide whether the playback is your subject. If it isn’t, plan to scrub or pause rather than play.
  2. Turn on System Audio before you start, so the clip’s sound is captured at the source.
  3. Play from a local file if possible; avoid streaming and avoid anything DRM-protected.
  4. Keep playback segments to a few seconds each.
  5. Close other apps — this is one of the genuinely demanding things you can record.

Afterward, in the editor:

  1. Trim the playback down to the part that earns its place.
  2. Duck or mute the system audio track under your narration.
  3. Add a zoom or a text overlay if the viewer needs to know where to look.
  4. Export MP4 at the resolution your content actually is, not the largest one offered.

Recording a video playing on your screen will never look as good as the video does. Once you stop trying to make it, and start treating the playback as something to frame and narrate rather than reproduce, the recordings get noticeably better. If you want to try the workflow above, download Recorded for macOS or Windows.