Send screeners you can trace
Every recipient gets their own copy, marked invisibly. If one leaks, the file itself tells you whose it was.
The problem with sending a screener
A screener goes to dozens of people — festival programmers, distributors, sales agents, press, cast and crew. Every one of them is someone you chose to trust, and every one of them is a point of failure. When a cut turns up somewhere it should not be, the usual position is that it could have been any of them.
Burned-in visible watermarks are the traditional answer, and they work, at a price: they are ugly, they distract the exact people whose job is to judge the film on its merits, and they are the first thing a determined leaker crops out.
How per-recipient watermarking works
Instead of one marked master, you issue one file per recipient. Each carries a different invisible identifier woven into the image data itself — not the filename, not the metadata, not a header that a re-encode would strip. Nobody can see it, including the recipient, and no two copies look different from each other.
When a file surfaces, you run it back through the detector. It recovers the identifier and tells you which copy you are holding. The evidence travels inside the picture, so it arrives with the leak.
What it survives
A leaked screener is rarely a pristine file. It has been recompressed, resized, screenshotted, filmed off a monitor, or passed through a platform that re-encodes everything it touches. A mark that only survives a clean copy is not much use.
Blurry publishes the actual numbers rather than adjectives. One photograph, one invisible ID, 12 attacks — recompression to low quality, resizing down to a quarter scale, screenshots, social re-uploads, rotation, and crops that throw away three quarters of the frame. The AI watermark came back from 12 of 12, and every confidence figure is real decoder output on the exact file shown beside it.
You can see the whole thing, attack by attack, on the robustness demo — including where it eventually fails, because a page that only showed successes would not be worth reading.
Video, and what it costs
Screeners are video, so video watermarking is the part that matters. It is included in the Premium tier at $20/month, which also unlocks the AI method — the one built to survive being handled. The free tier exists so you can satisfy yourself the thing works before paying for it.
There is no per-recipient licensing charge and no enterprise sales call. Full pricing is on the home page.
What this does not do
It does not stop a leak. Nothing does. Someone determined to put your film on the internet can still do it, and any watermark can be destroyed if the file is degraded far enough — at which point what is circulating is a bad copy of your film, which is its own kind of answer.
What it changes is the arithmetic. A recipient who knows their copy is individually identifiable behaves differently from one who believes they are anonymous in a crowd of fifty. Most screener leaks are not sophisticated attacks; they are someone forwarding a file because they assumed it could not be traced.
Try it before you trust it
Download a watermarked frame, attack it however you like, and see whether the ID still comes back.