Camera AirPods code shows the capture model: 0.1MP background frames, and an indicator light

MacRumors forum member mactracker found a hidden macOS framework called AccessorySensorManager in the Tahoe 26.7 release candidate, the same build that accidentally shipped a demo video of camera-equipped AirPods ^[2][1]. The framework handles sensor data from AirPods connected to a Mac, which means the always-on capture model that privacy reviewers and accessory developers have been speculating about is now partly written down ^[2].

The design is stereo. According to the code as read by mactracker, the left and right buds each carry a camera sensor and return synchronised RGB stills of up to 1 megapixel with matching frame IDs, giving Visual Intelligence the same scene from two angles ^[3]. No infrared sensor is mentioned, though the code does not rule one out ^[4]. There is no video path: the cameras produce internal still frames for Visual Intelligence to process rather than photos or recordings ^[5]. In active mode the framework either requests one synced pair or supplies a capture rate and receives pairs continuously ^[6].

The two resolution tiers matter more than the headline megapixel figure. Active mode, which mactracker expects to be user-triggered via Siri, captures 640x640 and returns a processed 1024x1024 frame, which mactracker describes as "about 0.4 megapixels of sampled image data, and a little over 1 megapixel for the output frame" ^[7][8]. Passive mode captures 320x320, with a 320x320 or 512x512 output ^[9]. That passive frame is 102,400 pixels, about a quarter of the sampled area in active mode ^[24]. mactracker adds that these are capture-specific resolutions and the physical sensor may be larger ^[10].

Passive mode is the part worth reading closely. The framework references contextual conditions including nearby speech, changes in the surrounding audio, posture changes, head rotation, and movement outside a defined area, and the code does not make clear whether every condition triggers a capture ^[11]. The engineering around a camera strapped to a moving head is also visible: lens distortion correction, dual-camera calibration, camera-to-motion-sensor calibration, and rejection of frames ruined by excessive movement or an occluded view ^[12]. Some work happens on the buds, described in the code as "peripheral inference," including detecting whether a person is in view ^[13]. The framework can also remotely control a hardware indicator and its brightness, which points to a capture light intended for the people around the wearer rather than the wearer ^[14].

The caveat is large. The buds in the video are a camera-equipped version of AirPods Pro 3, codenamed B790, shown reading a book title held up to them so Siri could save it for later ^[15][23]. Bloomberg's Mark Gurman reports Apple no longer plans to release that model and is working on a new generation codenamed B798, which makes it likely the framework documents the cancelled device ^[16][17]. Gurman says the camera AirPods effort remains on track for 2027 ^[19], is off the 2026 schedule while Apple resolves supply chain and software problems ^[20], and could arrive alongside the 20th-anniversary iPhone ^[18]. His own timeline has moved repeatedly: advanced testing in May, a delay to late 2027 in June, then a possible September launch in early August ^[21]. Gurman also notes privacy pushback, including critics calling the device "pervertpods," and says the cameras are low-resolution and meant to scan the environment for Siri context rather than shoot photos or video ^[22].

Watch whether B798 inherits the indicator light, the on-device person detection, and the passive tier at all, and whether Apple documents the passive trigger conditions publicly rather than leaving them to be reverse-engineered from a shipped framework ^[11][13][14].