YouTube air-gapping: why a filmed screen can't earn the camera label

Air-gapping is YouTube's name for pointing a camera at a screen that shows synthetic content so the recording picks up a genuine camera signature. YouTube's 'Captured with a camera' page explains why that does not work: the metadata behind the label is made by a third party such as a camera maker, and because the screen shows a modified image, the recording would not be eligible for the disclosure.

The page admits a residual risk and says camera manufacturers will keep developing detection measures to prevent air-gapping. Treat the label as proof of a verified capture chain, not as a stamp you can obtain by re-filming a monitor.

What the label requires

The disclosure appears in the 'How this content was made' section when YouTube detects secure C2PA metadata from a tool with built-in support for version 2.1 or higher, and the sound and visuals have not been edited. The table lists what breaks it.

If your clip came from Sume

A generated or trimmed clip from Sume is not a camera capture, and its docs do not describe adding C2PA data, so the camera label is not a goal for it. Sume's trim with precision: exact is a frame-accurate libx264 re-encode into a new artifact, which is exactly the kind of edit that ends a capture chain.

What applies instead is the other YouTube page: realistic AI-generated or meaningfully altered footage should be disclosed with the 'AI use' answer in Studio, and the GenAI disclosure page warns that creators who consistently skip it can face a manual label or removal from the YouTube Partner Program.

Steps

• If a clip is a real capture from a C2PA-capable camera and you want the label, upload the untouched original.
• If you must edit, do it in tools that preserve C2PA 2.1 or higher, and keep audio and visuals unaltered.
• If a clip is AI-generated and realistic, disclose it; do not try to launder it through a screen recording.
• Check the expanded description after upload to see which disclosure YouTube applied.

Why the rule is built this way

A capture label is only useful if viewers can trust it. If a screen recording of a generated scene could earn it, the label would stop meaning that a real camera saw a real scene. The page ties the label to metadata that a third party makes at the moment of capture, which is why a filmed monitor, whose image is already modified, is ruled out in the text itself.

For publishers this is a design constraint rather than a hurdle: decide per clip whether it is a capture or a generation, and follow that clip's own disclosure path.

Caveat

The page also says a missing 'Captured with a camera' label does not mean the content has modified audio or visuals. It only appears when a creator opted into C2PA capture.