The honest answer
Stop someone screen recording a video link.
In a browser, you cannot. Here is what actually works.
There is no capture-prevention lever on the web. The standards body that owns the closest thing says so in its own specification, and no browser vendor documents an opt-out for page content. So the useful question is not how to prevent a recording. It is how to make one traceable, short-lived and expensive to attempt.
Video hosting for filmmakers · 5 GB free · Paid plans from USD 9/month
Updated September 2026
The short answer
No web page can block a recording,
and no vendor documentation says otherwise.
You cannot stop someone screen recording a video link. Not on this platform, not on anyone else’s. The web has no capture-prevention lever to pull, and the standards body that owns the nearest thing to one states that in its own specification.
The API browsers use for encrypted playback is Encrypted Media Extensions. Its specification opens the discussion by ruling itself out: “This specification does not define a content protection or Digital Rights Management system”, adding that it defines a common API used to discover, select and interact with such systems. That is the W3C, on its own page, fetched 9 September 2026. Mozilla’s reference page for the same API describes interfaces for controlling the playback of content subject to a digital restrictions management scheme, then documents message exchange with a decryption module and key management. It mentions no HDCP, no output protection and no capture prevention anywhere.
There is exactly one output-protection lever a web application gets, and it is a question rather than a command. MediaKeys.getStatusForPolicy() checks whether a decryption module would allow presentation of encrypted media using given keys, and MDN states that the policy restrictions currently only include a restriction on the minimum supported HDCP version. The page can ask whether the path to the display is trusted at a given HDCP version. It gets back a status, and one of the possible answers, output-restricted, is documented as meaning the connection between the source and the output is not trusted. That is information about a display. It is not a hand on a recorder.
The other side of the story is just as plain. MDN’s Screen Capture API page documents the capture direction: its main method asks the user to select a screen or a portion of a screen to capture, and returns it as a media stream. The permission model asks the person doing the recording. Nothing on that page gives a page a way to exclude its own content from what is captured. Both halves were read on 9 September 2026, and both are linked in the sources below.
Every row is a direct reading of the primary source named in it, fetched 9 September 2026 and listed with its URL in the sources block at the foot of this page. No third-party article, benchmark or review site is used as a source anywhere on this page.
The conflation
DRM is a lock on decryption,
not a wall between a frame and a recorder.
Almost every product that promises to stop screen recording is selling you DRM and letting you assume the two are the same thing. They are not, and the gap between them is where the disappointment lives.
DRM answers one question: is this player, on this device, authorised to turn these bytes into pictures right now. It is genuinely useful. It stops a downloaded file from being playable elsewhere, it lets a licence expire, and it gives a distributor a defensible answer about how content moved. What it does not do, on a general-purpose computer, is stand between a decoded frame and a utility that reads the framebuffer. Once the frame exists on screen for a human to watch, it exists for software to sample.
Where a blackout does happen, it comes from the operating system, not from the web page or its DRM. Apple documents exactly that behaviour and its limit: the system does not include FairPlay-encrypted video when recording, mirroring or using AirPlay, and yet, in Apple’s own wording, although it blacks out FairPlay-encrypted video content it includes FairPlay-encrypted audio content. So even the strongest documented guarantee on the most controlled consumer platform hands the recorder a black picture with your full soundtrack under it. That is a real protection. It is not the protection people think they are buying.
Native applications
A difference of degree,
not of kind.
An installed application sits closer to the operating system than a page does, so it can be told things a page cannot. Being told is most of what it gets.
Apple’s UIScreen.isCaptured is documented as a Boolean indicating whether the system is actively cloning the screen to another destination, with a true value meaning the system is actively recording, mirroring or using AirPlay. Apple’s guidance on that page is to observe the property and optionally take an appropriate action, giving the example of a media app that stops playback and shows the viewer a dialog explaining the pause. The property is deprecated as of iOS, iPadOS and tvOS 27 in favour of UITraitCollection.sceneCaptureState, whose abstract describes it as representing whether a scene is currently being mirrored or recorded.
Read those two pages together and the shape is unmistakable. The current API is a reporting trait. It tells an application that a capture is under way; the application decides whether to pause, blank, warn or carry on. That is worth having, and it is a categorically different promise from making the pixels unrecordable.
The correction most articles on this subject get wrong
The constant usually cited as the macOS way to exclude a window from screen capture is NSWindow.SharingType.none. Apple’s own documentation page for that enumeration, fetched 9 September 2026, describes the none case as a legacy constant that macOS no longer uses, and marks readWrite as deprecated. No case description on that page mentions screen-capture exclusion.
If a vendor tells you their Mac player hides itself from recordings, that claim needs to rest on something other than this constant. Ask which mechanism, then read that mechanism’s own documentation page before you believe the sales copy. This is the single most repeated error in the published writing on this subject, and it is repeated because nobody opens the page.
The unfixable case
There is a phone in the room,
and it always wins.
Suppose every software path were closed. Suppose the frame never reached a capture utility, on any operating system, in any player. A person can still hold a phone up to the monitor and press record. The result is ugly, it is off-axis, it has a moiré pattern across the fine detail, and it is entirely good enough to spoil an embargo, reach a forum, or get quoted in a review before your release date.
Any security model that treats capture as preventable is theatre, because the last metre of the chain is a human eye and there is no API for that. Once you accept the camera in the room, the question changes usefully. You stop asking how to make a copy impossible and start asking what a copy costs the person who makes it.
What to do instead
Trade prevention for attribution,
deterrence and a short window.
The four controls below do not stop a recording. They make one attributable, they make it visibly attributable while the viewer is deciding, and they close the window in which it can happen. Run together, that is what a serious screener programme is.
A visible session watermark
The viewer's name, session and time carried across the frame, so the person watching knows that a copy of this exact viewing points at them. This is the control that changes behaviour, because it works on the decision rather than on the file. Frame.io documents its equivalent as displaying the viewer's name, email, IP address and date during playback.
An invisible per-person mark
A forensic watermark is an identifier the viewer cannot see, recovered later from a leaked copy. Frame.io describes its own as an invisible pixel that is not visible to the viewer, capturing session ID, user info and location data, and states plainly that it does not run detection itself: the customer obtains the leaked file and works with the watermark vendor. Treat recovery as conditional, never as a guarantee.
A short-lived link
A signed link that expires in minutes rather than days means a URL pasted into the wrong thread is usually dead before anyone opens it. Expiry does not prevent a leak during the review. It bounds the window in which one is possible, which is the only honest thing expiry ever does.
A view log you can audit
Who opened it, when, and from where. A log will not stop a recording either, but it turns an investigation from a guess into a shortlist, and it tells you within hours if a link reached someone you never sent it to.
The other half is not technical at all. One recipient per link, never one link for a distribution list. An NDA that names the consequence rather than gesturing at confidentiality, signed before the link is issued. An explicit no-forwarding term, so that passing the link on is itself the breach and you do not have to prove what happened after. And revocation as a scheduled task the day a review closes, verified the way you verify a backup, by opening your own link in a private window and confirming it is dead.
None of that is glamorous, and all of it outperforms any promise to block capture. Deterrence plus attribution plus a short window is a defensible programme. Prevention is a claim nobody can honour.
What the competition actually claims
The claim shrinks when you
reads it carefully.
Frame.io is the reference point most people reach for, so it is worth reading its documentation precisely rather than from memory. Its current V4 Content Security page, fetched 9 September 2026, documents the settings available: applying DRM to video assets across the account, controls for internal playback and share links, watermark and forensic-watermark toggles, and a fallback if forensic watermarking fails. It does not state what DRM prevents, and it says nothing at all about screen-recording prevention.
The screen-recording claim appears in one place: a legacy article describing Frame.io V3, which states that DRM safeguards content against piracy or leaks and prevents screengrabs and screen recordings. Read the next part of that article and the claim narrows sharply. It scopes that prevention to Safari on macOS, Edge on Windows, Chrome on ChromeOS, and the native Frame.io apps for iPhone, iPad and Apple TV, while separately noting that DRM playback itself is supported across Chrome, Firefox, Edge and Safari including mobile browsers. Prevention on a named list, playback on a broad one. Those are two different sentences and the marketing version tends to merge them.
We are not going to tell you whether that legacy list still applies to Frame.io V4, because Frame.io does not document it and we will not guess on their behalf. What is fair to take from it is the shape of every honest claim in this category: capture restrictions live in specific clients on specific operating systems, and the moment you hand someone a plain browser link, you are back at the top of this page.
Access control
Passwords, per-recipient invitations, expiry and revocation. Mature, well documented across the category, and genuinely effective against the most common leak, which is a forwarded link.
Attribution
Visible session marks and invisible forensic marks. These are the controls that change a viewer's behaviour and give an investigation something to work with. Recovery from a recompressed or re-recorded copy is conditional everywhere it is offered.
Capture prevention
The one thing everybody wants and nobody can sell you in a browser. Where it exists, it lives in a specific native client on a specific operating system, and it stops at the camera in the room.
Where our own line sits
A property of an application,
not of a link.
A master under Vault on uncompressed.io is native-app only. Web playback, share pages and browser downloads go dark, and the raw master never reaches a browser. Inside the macOS app the window is excluded from recordings, screenshots and Zoom or Teams shares, and it composites black. Every frame carries a visible session watermark with the viewer, the session and the time. Every frame also carries an invisible per-person forensic mark. The master is stored AES-256 encrypted at rest, links are signed and short-lived, downloads and export are off, and every open, play and capture attempt lands in a tamper-evident view log.
Now the boundary, in plain words, because blurring it is how this category loses trust. The screen-capture block is a property of the native application path. It is not a property of a browser link. A film you share as an ordinary uncompressed.io link plays in a browser, and every limit described at the top of this page applies to it in full, on our platform exactly as on any other. FairPlay DRM appears on our roadmap and has never shipped, so it protects nothing you send today.
What that buys, when the material warrants it, is a viewing your recipient must make a deliberate choice to circumvent, a copy that points back at one person if they do, and a log that tells you when to start looking. It does not buy a guarantee, and you should be suspicious of anyone selling one. The screener guide lines every control up against the leak it addresses, and the security page documents each switch on its own terms.
When you should not buy any of this
If your material is a commercial that airs in ten days, a corporate film, a wedding cut or a portfolio piece, none of this is for you. A private or password-protected link with downloads off and an expiry on it is proportionate, costs nothing and annoys nobody. Buy the heavy machinery when a leak would actually cost you a release, a festival slot or a distribution deal.
And weigh the friction honestly, because you are the one imposing it. Requiring a distributor, a programmer or a journalist to install an application before they can watch your film is a real cost, paid by them, on your behalf. Some will do it without comment for an unreleased feature. Some will ask you why, and for routine work they will be right to ask.
Questions
Frequently asked
Can a website stop someone screen recording a video?
No. The Screen Capture API documentation on MDN, fetched 9 September 2026, describes how a page asks the user for permission to capture a screen. It documents no mechanism for a page to exclude its own content from a capture. The consent model asks the person recording, not the page being recorded.
Does DRM stop screen recording?
DRM controls decryption and playback authorisation. The W3C Encrypted Media Extensions specification, fetched 9 September 2026, states that it does not define a content protection or Digital Rights Management system, and defines an API for interacting with one. Nothing in it defines capture prevention. Any platform-level blackout comes from the operating system, not from the web page.
What about HDCP? Does that block a recorder?
HDCP concerns the link between a machine and a display output. The only related web API is MediaKeys.getStatusForPolicy(), and MDN states that its policy restrictions currently only include a restriction on the minimum supported HDCP version. It is a query about whether keys would be usable, not a block. A page can ask about a display. It cannot reach a recording utility.
Do native apps do better than a browser link?
Yes, but it is a difference of degree, not of kind. Apple documents capture detection through UIScreen.isCaptured, deprecated from iOS 27 in favour of UITraitCollection.sceneCaptureState, and states that FairPlay-encrypted video is blacked out during recording while FairPlay-encrypted audio is still included. The app is told a capture is happening and decides what to do about it.
Is NSWindow.SharingType.none the way to exclude a Mac window from capture?
Apple's own documentation for that enumeration, fetched 9 September 2026, describes the none case as a legacy constant that macOS no longer uses, and marks readWrite as deprecated. No case description on that page mentions screen-capture exclusion. It is the constant most often cited in articles about this subject, and it is not the answer.
Does the uncompressed.io Vault block screen recording?
In the macOS app, yes: the window is excluded from recordings, screenshots and Zoom or Teams shares, and it composites black. That is a property of the native application path. It is not a property of a browser link. A film shared as an ordinary link plays in a browser, and everything at the top of this page applies to it. FairPlay DRM is on our roadmap and is not shipped.
Should I make my reviewers install an app?
Only if the material justifies it. A native-app requirement is real friction to impose on a distributor, a programmer or a journalist, and some of them will push back. For a client review of a commercial that ships next week, a private link, downloads off and an expiry are proportionate. For an unreleased feature under embargo, the friction is the point.
Attribution beats a promise nobody can keep
Vault masters play in the macOS app only, with capture blocked, a visible session watermark, an invisible per-person forensic mark, AES-256 encryption at rest, short-lived links and a tamper-evident view log. The capture block belongs to the native app, not to a browser link, and this page says so on purpose.
Sources
- 1.W3C, Encrypted Media Extensions specification (fetched 9 September 2026; the specification's own statement that it does not define a content protection or DRM system)
- 2.MDN, Encrypted Media Extensions API (fetched 9 September 2026; key exchange and decryption module messaging only)
- 3.MDN, MediaKeys.getStatusForPolicy() (fetched 9 September 2026; minimum HDCP version is the only documented policy restriction)
- 4.MDN, Screen Capture API (fetched 9 September 2026; documents capture and its permission model, no opt-out for page content)
- 5.Apple Developer, UIScreen isCaptured (fetched 9 September 2026; capture detection, the FairPlay video blackout, and the audio exception)
- 6.Apple Developer, UITraitCollection sceneCaptureState (fetched 9 September 2026; the replacement trait from iOS 27)
- 7.Apple Developer, NSWindow.SharingType (fetched 9 September 2026; the none case is documented as a legacy constant macOS no longer uses)
- 8.Frame.io Knowledge Center, Content Security (fetched 9 September 2026; documents the DRM and watermark settings, states nothing about screen-recording prevention)
- 9.Frame.io Knowledge Center, Digital Rights Management (Legacy) (fetched 9 September 2026; the screengrab and screen-recording claim, scoped to a named browser and operating-system list plus native apps)
- 10.Frame.io Knowledge Center, Watermarking in V4 (fetched 9 September 2026; what a session-based watermark displays, and the tiers it is sold on)
- 11.Frame.io Knowledge Center, Forensic Watermarking (fetched 9 September 2026; invisible mark, Enterprise Prime only, detection handled through the watermark vendor)
