Watermarking is the security control people skip until the morning after a leak. Frame.io's own blog reopened the topic in February 2026 with an op-ed titled "Watermarking: You Don't Think About It - Until You Need It" - and they are right about the timing problem, which is why this post is the practical decision guide: which type of watermark protects which threat, how visible to make it, and when the honest answer is that a watermark is theater. (The persuasive case - why every review cut should carry one - already lives in why watermark video reviews; this post assumes you are convinced.)
The three types, matched to threats
Static visible watermark (logo or "PREVIEW" text). Deters casual screenshot reuse and marks the file as unfinished. Protects against: a reviewer pitching your cut as their own, an unfinished cut leaking and being mistaken for final. Does not protect against: identifying who leaked, because every viewer sees the same mark.
Per-viewer dynamic watermark (viewer name, email, IP, timestamp burned in). The one that actually answers "who leaked it". Each reviewer's copy carries their identity, so a screenshot that escapes is traceable to the account that generated it. Dropbox Replay advertises exactly this on its review product - viewer IP/date/time tracking on watermarks - and reviewer-name burn-in is standard on serious review tools. This is the default for pre-release commercial work, festival submissions, and anything under embargo.
Forensic watermarking. Imperceptible marks embedded in the stream itself, robust against re-encoding and crops, readable with the vendor's detector. This is studio-dailies territory (it is how feature-screening leaks get traced) and it is a specialist capability - most review platforms do visible per-viewer marks, not forensic ones. If a client demands forensic marking, that is a dailies-security conversation with specialist vendors, not a review-tool checkbox.
| Type | What it stops | What it does not | Typical use |
|---|---|---|---|
| Static visible | Casual reuse, "unfinished" confusion | Leak tracing | General client rounds |
| Per-viewer dynamic | Leaks (traceable to the account) | Re-encode removal (it is visible) | Pre-release, embargoed, cutdowns before air |
| Forensic | Leaks surviving re-encode/crop | Nothing - but requires specialist tooling | Studio dailies, high-security screenings |
Placement and opacity that reviewers tolerate
A watermark that makes the cut unjudgeable fails its own purpose - reviewers need to evaluate color and composition through it. Working practice: 30-50 percent opacity, positioned where it overlaps content but not faces or key supers, and moving (corner-cycling) marks if you expect deliberate cropping - a static corner mark can be cropped out, a mark that changes position every 20-30 seconds cannot be removed without visible damage. For per-viewer marks, name plus timestamp in a lower corner is enough; full-screen diagonal tiling is for high-security contexts where judgeability matters less than deterrence.
When watermarking is overkill
Internal rounds, cuts already published, and reviewers who are contractually bound and technically locked down (SSO, managed devices) gain little from visible marks - and a heavy watermark on a color-sensitive round (the graded commercial, the film pass) actively damages the review. Match the watermark weight to the leak consequence: a restaurant's website video leaking early costs little; an unreleased film leaking costs everything. The first gets a light corner mark; the second gets per-viewer marks, passwords, expiry, downloads off, and possibly forensic tooling.
Doing it: in the tool or in the export
Two implementation layers exist. Review-tool watermarks apply at the link level - every viewer of that link sees the mark, ideally their own identity - and nothing about your source files changes. NLE-side watermarks (burned into the export, e.g. Resolve's Text+ approach covered in how to watermark review exports in Resolve) travel with the file forever, which is right for files that leave the platform: festival submissions, press kits, the client's own distribution copies. The rule of thumb: link-level marks for review rounds, file-level marks for files that escape your control.
FAQ
What watermark type should I use for client review videos?
Per-viewer dynamic watermarking (viewer name and timestamp burned into the frame) for anything pre-release or embargoed, a static light mark for ordinary client rounds. Pair either with a link password, expiry, and downloads-off - watermarking alone is one control, not a security system.
Can a watermark identify who leaked a video?
Only a per-viewer watermark can - each reviewer's stream carries their identity, so a leaked screenshot or recording traces back to the account that generated it. Static marks identical for every viewer cannot identify anyone.
How visible should a review watermark be?
Working practice: 30-50 percent opacity, placed to overlap footage but not faces or supers, and position-cycling every 20-30 seconds if you expect cropping attempts. The mark must not make color and composition unjudgeable - a review watermark that breaks the review defeats itself.
What is forensic watermarking and do I need it?
Forensic watermarking embeds imperceptible, re-encode-resistant marks readable with specialist detection - it is standard for studio dailies and high-security screenings. Most client work is served by visible per-viewer marks; step up to forensic only when the leak consequence justifies specialist tooling.