Privacy masks

An administrator uses Privacy masks to cover areas such as a neighboring window, keypad, or screen. Select the camera, choose Draw mask, place the corners, close the shape, give it a name, and save. Check the resulting picture and every viewing or export path the site intends to use.

Each privacy form has its own Save button. Drawing or editing a mask stays local until you save; Cancel discards those edits. To turn a mask on or off, choose Edit, change Enable mask, and select Save. The saved mask continues to apply while its edits are unsaved.

Masks differ from detection zones: zones define areas for rules, while masks conceal picture content. The camera's own configuration is not changed, and other systems connected directly to that camera are unaffected.

Fixed and moving cameras

A fixed mask stays at the same place in the picture. On a supported PTZ camera, Follow the camera binds the mask to the camera's current pointing direction. This needs reliable camera geometry; a default or poor calibration can drift. The mask is bound only if the picture you drew it on was taken after the camera settled from its last move: a PTZ camera reports where it is going before its picture gets there. If the camera moved after that picture, Save says so; check the corners against the current picture and save again. The same applies to each placement while training.

Use the geometry controls to calibrate, or Train by hand to place the mask after a sequence of camera movements. Check positions between the training samples as well. During training, camera triggers and alerts are paused and the administrator uses a dedicated preview with a translucent training mask.

While the camera moves controls whether the picture is blanked or masks continue following the movement. Position reporting can lag the video. Moves made outside EverNVR are detected by position polling, up to half a second later. Select Save in this form to apply a new movement policy. Manually entered camera geometry also takes effect only after its form is saved. Calibration and training remain explicit actions with their own instructions and confirmations. The same geometry is what the site plan measures a PTZ camera's heading through, and what auto tracking steers by, so a camera with only the default geometry cannot be aimed at a point on the plan or follow what it detects.

If the required shaped-mask runtime is missing, the picture is withheld rather than shown without its mask. Resolve that warning before relying on the view. The runtime is the same detector installation that Settings → Detection reports, so a recorder whose detector reads Ready should not show it. Release 0.1.43 showed the warning on every installed recorder regardless; 0.1.44 corrects that.

Stored media

Recordings on disk preserve the camera's original video packets. Live snapshots, detector input, event images, and stills extracted from recordings have masks applied to their pixels. Every clip pulled from a masked camera, whether played on the timeline, downloaded, or fetched by another system, is re-encoded with the masks burned in; it begins at the moment asked for rather than at the keyframe before it, and it costs an encoder for as long as it plays. The shapes are burned as drawn: a clip from a release before 0.1.49 covered each shaped mask with its surrounding rectangle.

Live video in the browser is the one exception. It is the camera's own packets, and the page draws the mask shapes over them, because re-encoding every live stream would undo the cost model that lets one recorder carry many cameras. That overlay exists only in the browser, so a masked camera's live video socket is refused to API tokens, which are told to use stream.mjpg and frame.jpg instead; both have the masks applied to their pixels. A PTZ camera whose position is not yet known shows a blank picture rather than a mask drawn where the camera once pointed.

Until a camera with masks that follow it reports where it is pointing, for example while it restarts, every masked picture of it (snapshots, frame.jpg, stream.mjpg, and detector input) is blacked out completely rather than withheld, so snapshots still succeed. Meanwhile the camera reads as degraded with Waiting for the camera to report where it is pointing; privacy masks need its position. The masks apply as normal once the position arrives. Some cameras take a minute or more to report it after a restart.

Event thumbnails and still images from recordings preserve the mask's polygon shape, including when cropped around an event. Existing saved thumbnails keep the masking applied when they were generated.

Protect raw media and backups with host and storage permissions: a mask changes what leaves the recorder, not what is on its disk. Test the intended viewing and export paths, and review files before sharing them.

See security and data for account and storage boundaries.