Skip to content
VVideoTextRemove

Tutorials, demos and capture

Remove text from a screen recording

A screen recording is the best case this tool has, and it is worth understanding why. In ordinary video the background moves, so every repaired pixel has to be inferred from a moving neighbourhood. In a screen recording the background is usually a static window: the same pixels, frame after frame. Removing text from it is closer to copying than to inventing.

The catch is that usually is load-bearing. Scroll a page, play a video inside the recording, or let something animate behind the text, and the easy case disappears.

Before — burned inAfter — reconstructed

Drag the handle — the caption band is rebuilt from neighbouring frames, not covered with a blur box.

Doing it for this case

  1. 1

    Separate interface from overlay before marking anything. Menu labels usually belong in the final cut.

  2. 2

    Box each overlay separately in a single job rather than one large region.

  3. 3

    If the page scrolls under the text, treat that section like ordinary footage and expect the repair to soften.

Why static UI is the easy case

When the picture behind the text does not change, the model has an unlimited supply of clean copies of it — the frames before the overlay appeared and the frames after it left. A tooltip over a toolbar, a notification over a desktop, a caption burned into a demo: all of them sit over pixels that exist unaltered elsewhere in the clip.

This is the one scenario where the result is often indistinguishable from the original at 100%, provided the box is tight. Every pixel of extra area is invented rather than copied, and invention is where artefacts come from.

  • Text over a gradient or a photo wallpaper is still fine — static is static
  • Keep the box tight: a generous box on a static background is the most common self-inflicted artefact
  • Run the job on the shortest section that contains the overlay if you only need one segment

What breaks it

Three things. Scrolling, where the content under the text moves so the clean copy is in a different place in every frame. Video playing inside the recording, which is the same problem plus a second compression pass. And anything animated — spinners, progress bars, cursor trails — which behaves like ordinary footage with the same failure modes.

The cursor deserves its own note. It is not distinguished from background, so a cursor that crosses the box becomes part of what the model samples, and it gets smeared into the repair.

  • Scrolling page: split the job at the point where the text stops moving with the page
  • Blinking caret: include its full blink cycle, or it returns as a flicker
  • Hover states: box the region only for the frames where the overlay is actually there

Deciding what counts as text

A tutorial recording is not one overlay but several. A title card, tooltips, menu labels, and a burned-in subtitle track all live in the same frame, and they need different decisions. Menu labels are part of the interface and should stay; a tooltip you no longer want is a small, easy job; a subtitle band across the bottom is a wide, thin strip with the same geometry as any other caption.

Remove each overlay as its own box in one job. One large box around the whole lower half of the screen turns three easy repairs into one hard one.

Tested, not assumed

A static background is what our foliage test approximates — 28.3 dB, with the repaired band the only soft area.

Case A is a moving camera over foliage: not static, but consistent, which is the property that matters. It scored 28.3 dB inside the repair against 37.4 dB outside it. A screen recording with a genuinely static window has more usable information than that clip does, so the repair should be at least as clean — we have not measured a screen recording, so we are not putting a number on it.

Where it breaks

  • Scrolling content or video playing inside the recording behaves like moving footage: expect softening.
  • The cursor is not separated from background; crossing the box smears it into the repair.
  • UI text at small sizes with anti-aliasing leaves a faint halo if the box is a pixel short.
See the footage and the measurements

Frequently asked questions

Can I remove subtitles burned into a screen recording?+

Yes. A subtitle band in a recording is the same shape as any other caption band, and because the UI behind it is static the result is usually cleaner than on camera footage.

What about a trial-software watermark?+

Yes if it sits in one place. A static watermark over a static interface is the easiest job this tool has — box it once and the repair copies the surrounding pixels.

Can it remove a notification popup?+

Yes, and the frames where the popup is absent are the clean source for the repair. Box the popup only, not the area around it.

Does the recording quality matter?+

Yes. Heavy compression puts blocky artefacts around text, and those artefacts sit outside the text itself. A higher-bitrate capture gives the repair cleaner edges to work from.

Will the layout shift after removal?+

No. The frame keeps its dimensions and nothing is rescaled, so the interface stays exactly where it was.

Related tools

Try it on your own clip

Sign in and get 1 free credit — enough for a one-minute video. No card required.

Open the editor