Guides

How to Blur Sensitive Data in a Screen Recording

What actually leaks in a screen recording, how to hide it before, during and after capture, and the three places people forget to look until it is public.

The stepvideo Team

stepvideo editorial

Published
2 October 2026
Reading time
8 min read
A browser form where most field rows stay crisp while three rows holding personal data are replaced by coarse pixelated blocks

Start by not recording it

Every tool in this category sells blurring, so every article about it starts with blurring. That is the wrong order. The cheapest way to keep a customer's name out of a tutorial is to be logged into an account where no customer exists, and a seeded demo account with invented names, invented invoice numbers and an invented domain removes nearly every decision you would otherwise make under time pressure.

This matters because blurring is a per-recording tax that you pay every time, and taxes get skipped when someone is in a hurry. A demo account is set up once. If your team records the same handful of processes repeatedly, the time spent seeding that account pays for itself inside a month and removes a whole class of mistake permanently.

What actually leaks

When people picture a leak they picture a credit card number on screen. In practice the embarrassing ones are duller and come from places nobody checked, because the person recording was concentrating on the task rather than the frame.

  • The browser frame itself. Bookmarks bar, open tabs, a tab title containing a customer's company name, a profile avatar, an extension badge showing an unread count.
  • Autofill. Click a field and the browser helpfully offers your real email, or a previous customer's address, in a dropdown that lasts two frames and is perfectly legible when paused.
  • The URL. Account identifiers, session tokens and query parameters sit in the address bar for the whole recording.
  • Notifications. A desktop or in-app notification arriving mid-take has shown more private messages than any mishandled form field.
  • What you said. Narration is not pixels and no blur tool touches it. Saying "so for Acme's account we would…" leaks a customer whose name never appeared on screen.
  • The results of a search. Typing a test query into a real system returns real rows, which is the single most common way customer lists end up in tutorials.

Hide during capture, not after

There is a meaningful difference between hiding something while recording and blurring it afterwards, and most comparisons skip it. If the region is covered at capture time, the underlying pixels are never written to the file. If you blur in an editor afterwards, the original frames still exist in the source recording, and the protection only applies to what you export.

For a tutorial that stays inside your company that distinction may not matter. For anything leaving the building, or anything you would have to explain to a security reviewer, it matters a lot — "we blurred it in the export" and "it was never captured" are different statements about where your customer's data currently lives.

The two places people forget

The first is the written guide. Any tool that produces step-by-step documentation alongside the video is also extracting screenshots, and those screenshots are generated from the same frames. A region hidden in the video is usually hidden in them too — but a region you fixed afterwards, in a video editor, certainly is not. Open the guide and read it as a separate document, because that is how it will be read.

The second is the transcript and captions. If narration is transcribed, anything you said aloud is now also written down, searchable, and often embedded in the exported file or published alongside it. Checking the video and forgetting the caption track is a common and complete failure of the whole exercise.

When you cannot avoid real data

Support is the case where none of the advice above applies cleanly. You are recording a real account because that is the account with the problem, and a seeded demo would not reproduce it. Here the goal shifts from prevention to containment: capture the narrowest scope that shows the issue, keep the recording out of anywhere indexed or public, and decide at the moment you create it when it will be deleted.

That last point is the one teams skip, because storage is cheap and deleting feels like losing evidence. But a support recording has a useful life measured in days, while the customer data inside it stays sensitive indefinitely. A recording that resolved a ticket in March and still sits in a shared drive in December is pure liability with no remaining value — set a retention rule once, at the folder or workspace level, rather than relying on anyone to tidy up later.

It is also worth agreeing internally on what gets recorded at all. Screens showing payment details, identity documents or health information usually warrant a written description rather than a capture, even when a recording would be faster to produce. The time saved is not worth the question you would have to answer if the file ever moved.

A checklist that takes two minutes

  1. 1Record in a demo account wherever the process allows it, and stop reading here.
  2. 2Before pressing record: close unrelated tabs, hide the bookmarks bar, turn on do-not-disturb, and sign out of anything personal.
  3. 3During capture, cover the regions you know will show real data rather than planning to fix them later.
  4. 4After exporting, scrub the video at speed and pause at every screen change — leaks appear on transitions, not in the middle of steady shots.
  5. 5Open the written guide and the captions separately and read both as a stranger would.
  6. 6Only then share it, and prefer an unlisted link over a public one for anything containing real screens at all.

If you want to try the during-capture approach, stepvideo lets you hide an element or drag a region while recording, and the same hidden areas carry into the written guide's screenshots. The honest caveats are above: it is a paid-plan feature at export, it is manual rather than automatic, and it only covers a browser tab. One caveat worth planning around: because it is paid-only at both ends, the 10-day trial is not where to evaluate it — the hide tool is refused on a free workspace, so you would be testing its absence. And if you only ever record a seeded demo account, you will not need it at all.

Related: how to record your screen in Chrome covers capture scope, and screen recording for customer support covers the case where real customer screens are unavoidable.

Frequently asked questions

Can you blur something after a screen recording is finished?

Yes, in a video editor, but understand what it does. The exported file hides the region while the original recording still contains it in full. For anything leaving your organisation, hiding during capture is the stronger position because the data was never written to disk.

Does blurring protect against someone de-blurring the video?

It depends which kind of cover it is, and most articles flatten this. A blur is a linear convolution: the detail is attenuated rather than destroyed, and high-contrast text at a modest radius can be partly recovered by sharpening. A solid block, or a mosaic that downscales the region to a few pixels, averages the information away and cannot be undone. Prefer the second for anything that matters, and treat a soft blur over a password field as discouragement rather than protection.

What about data I mention out loud?

No blurring tool touches narration, and if the audio is transcribed into captions it becomes searchable text as well. Decide before recording how you will refer to customers — a consistent invented name is easier than remembering to avoid a real one mid-sentence.

Is recording a customer's real screen ever acceptable?

Sometimes it is unavoidable, in support especially. Where it is, prefer the narrowest capture scope available, share through an unlisted link rather than a public page, and delete the recording on a schedule rather than keeping it indefinitely because storage is cheap.

Do these tools automatically detect personal data?

Treat any such claim with suspicion, including ours. Detection is useful as a prompt but should never be your only control, because the cost of one miss is unbounded while the cost of placing the cover yourself is a few seconds per recording.

The stepvideo Team

stepvideo editorial

We build stepvideo. Record a workflow in Chrome once and get back a cut, zoomed, narrated tutorial plus a written guide from the same take. Try it free.

Keep reading