WWDC Notes

Introducing Multi-Camera Capture for iOS

Show Apple's description.
In AVCapture on iOS 13 it is now possible to simultaneously capture photos and video from multiple cameras on iPhone XS, iPhone XS Max, iPhone XR, and the latest iPad Pro. It is also possible to configure the multiple microphones on the device to shape the sound that is captured. Learn how to leverage these powerful capabilities to bring creative new features like picture-in-picture and spatial audio to your camera apps. Gain a deeper understanding of the performance considerations that may influence your app design.
  • Supported on iPhone XS, XS Max, XR, iPad Pro 3rd Gen
  • AVCaptureMultiCamSession
    • Multiple AVCaptureDeiceInputs, AVCaptureDeiceOutputs of the same type, AVCaptureVideoPreviewLayers
    • Don't use implicit connection forming but use addInputWithNoConnections, addOutputWithNoConnections or AVCaptureVideoPreviewLayer.setSessionWithNoConnection
  • AVCaptureSession is still the way to got for single cam session
    • Simultaneous photo shooting, movie recording, barcode scanning, etc.
  • Limitiations
    • only one input per camera in a session
    • connecting one camera to multiple video data outouts is not possible
    • no presets supported on session since different cams might run with different qualities
    • multi-cam session has hardwareCost reporting. Session runnable when 0 <= cost <= 1
    • lower cost by lower resolution, choose binned format, deviceInput.videoMinFrameDurationOverride = CMTimeMake(1, 30) to set max framerate override /* 30 FPS */
    • lower system pressure like temperature, power demands, infrared projector temperature by
      • lowering frame rates, throttle GPU/CPU processing code, disable one camera
    • run indefinitely with multiSession.systemPressureCost < 1.0; device shutdown with cost > 3.0
    • iOS can run only one session at a time (with mutli cams though)
  • Virtual Camera is the new name wor software cameras like True-Tone- or Dual-Camera
    • device.isVirtualDevice - get its physical devices by device.constituentDevices for e.g. synchronized camera streaming
    • AVCaptureDataOutputSynchronizer ensures you get two outputs in one callback
    • virtual devices have secret ports so you can get 2 streams - you need to explicitly query them
  • Dual Camera Hography Aids
    • CameraIntrinsics (Optical center / focal length) and CameraExtrinsics (rotation matrix / translation vector for both wide- and tele cameras)
  • Multi-Microphone capture
    • By default front cam uses front mic and back cam uses back mic
    • actually front mic and back mic are a lie since different devices have multiple mics but not explicitly fron/back ones. This is achieved by Microphone Beam Forming - done automatically by CoreAudio micInput.ports(for: .audio, sourceDeviceType: micDevice.deviceType, sourceDevicePosition: .front).first
    • Beam forming only works with built-in mics
    • audio can be arbitrarily configured by creating custom AVAudioSession

This note was originally published at https://github.com/Blackjacx/WWDC

Missing anything? Corrections? Contributions are welcome 😃

Written by


Stefan Herold

Stefan Herold is an iOS Developer a.d. 2009 • iOS / OSX Enthusiast • WWDC20 • Now: @ioki_mobility • Passionate Mountainbiker