Skip to main content

Media capture, audio focus, and media sessions

Media features coordinate scarce hardware, lifecycle, privacy, output files, and other apps. A successful API call is only one part of correctness.

CameraX use cases​

  • Preview: camera frames displayed to the user.
  • ImageCapture: high-quality still capture.
  • ImageAnalysis: frames delivered to analysis code.
  • VideoCapture: encoded recording pipeline.

Bind compatible use cases to a lifecycle owner. Hardware capability affects which combinations, resolutions, frame rates, and concurrent cameras are available.

cameraProvider.bindToLifecycle(
    lifecycleOwner,
    CameraSelector.DEFAULT_BACK_CAMERA,
    preview,
    imageCapture,
    imageAnalysis,
)

Analysis must release each frame:

imageAnalysis.setAnalyzer(executor) { imageProxy ->
    try {
        barcodeScanner.process(imageProxy)
    } finally {
        imageProxy.close() // otherwise pipeline stalls
    }
}

Choose a backpressure strategy. If analysis is slower than camera input, “keep only latest” suits scanning; blocking the producer may suit frame-complete scientific capture but raises latency.

Capture state machine​

Idle -> Preparing -> Recording -> Stopping -> Finalizing -> Ready
                  ↘ Error       ↗

Handle permission revoked, app backgrounded, rotation, storage full, microphone unavailable, camera disconnected, and a zero-byte/partial result. Publish shared media only after finalization.

Audio focus​

Before audible playback, request focus with audio attributes describing content and usage. Respond to focus loss:

  • transient loss: pause, then optionally resume only if your policy/user intent permits;
  • ducking: lower volume when appropriate (not for spoken-word content where intelligibility suffers);
  • permanent loss: pause/stop and abandon focus;
  • delayed gain: remain paused until granted.

Do not automatically resume after every transient loss if the user paused during interruption.

MediaSession is the external control contract​

A session exposes playback metadata, queue, commands, and state to lock screen, Bluetooth, Assistant, Android Auto, and other controllers. Validate custom commands/controller access; keep standard commands consistent with actual player capability.

Background behavior​

Background playback belongs in MediaSessionService with appropriate foreground behavior. Background capture is much more privacy-sensitive and restricted: start from an allowed user-visible flow, declare correct foreground-service type/permissions where applicable, show clear ongoing indication, and stop when the task ends.

Never fake capture success

Camera/video callbacks can report asynchronous finalization errors. Confirm the output is finalized and readable before storing READY or showing “Saved.”

Sources​

Exercise​

Design a video-note recorder that supports rotation, 10-minute files, Bluetooth audio interruption, storage exhaustion, backgrounding, upload-after-capture, and deletion during upload.