iOS 15 Background Blur in Third-Party Video Apps

iOS 15 introduces system-level background blur—known as Portrait mode—for video calls, allowing third-party communication apps like Zoom, Microsoft Teams, WhatsApp, and Google Meet to utilize Apple's native depth-of-field effect. Rather than relying on individual app developers to code custom background segmentation, iOS 15 handles the blur directly through the operating system and Apple's hardware engine, accessible globally through the Control Center.

How the Feature Works at the System Level

Prior to iOS 15, background blurring in non-Apple apps required each developer to build or license proprietary computer vision models. These in-app solutions varied drastically in quality, consumed significant battery life, and often caused thermal throttling.

With iOS 15, Apple introduced a unified API within the AVFoundation framework. When a communication app captures a video stream using the camera, the operating system can intercept the feed and apply a hardware-accelerated Portrait effect directly at the driver level before the frames are transmitted over the network.

How Users Enable It in Non-Apple Apps

Unlike native in-app filters, this background blur is controlled via the iOS system interface:

  1. Open any third-party video communication app (e.g., Zoom, WhatsApp, Instagram, or Teams).
  2. Start or join a video call to activate the camera.
  3. Swipe down from the top-right corner of the screen to open the Control Center.
  4. Tap the Video Effects tile located in the upper-left corner of the Control Center.
  5. Tap Portrait to toggle the background blur effect on or off.

Once turned on, the setting persists for that specific application across future calls until manually disabled.

Hardware Requirements and Processing

The background blur feature relies heavily on the Apple Neural Engine (ANE) rather than traditional CPU or GPU compute shader methods.

  • Required Hardware: The feature requires a device equipped with the A12 Bionic chip or later (iPhone XS, iPhone XR, and newer models). Devices with older processors cannot run the model in real time without unacceptable latency.
  • Depth Estimation: On devices with TrueDepth camera arrays (Face ID models), the system combines raw depth data with machine learning segmentation masks. On devices with standard front cameras (such as the iPhone SE 2nd and 3rd generation), the ANE runs a real-time semantic segmentation algorithm entirely via software analysis of the camera feed.
  • Performance Gains: Because the processing is offloaded to the dedicated Neural Engine, non-Apple apps experience lower CPU usage, reduced heat generation, and better battery efficiency compared to their native software blur algorithms.

Developer Integration

For developers of non-Apple communication apps, adopting the feature requires minimal effort. By default, iOS 15 automatically opts in any app that uses standard system camera capture sessions (AVCaptureSession) for VoIP or video calling categories. Developers can query the system to see if the effect is currently active or programmatically opt out if their application requires an unaltered raw camera stream.