Noise Cancellation

Noise Cancellation capabilities of our iOS Video SDK can be enabled by installing our NoiseCancellation package. Under the hood, this package uses the technology developed by krisp.ai.

Installation

Add the SDK to your project

To add StreamVideo SDK, open Xcode and follow these steps:

• In Xcode, go to File -> "Add Packages..."

• Paste the URL https://github.com/GetStream/stream-video-noise-cancellation-swift

• In the option "Dependency Rule" choose "Up to the Next Major Version," in the single text input next to it. The field will be populated automatically with the latest version.

• Choose "Add Package" and wait for the dialog to complete.

• Select StreamVideoNoiseCancellation and add it in your project.

Integration

Our iOS SDK provides a utility component that makes the integration smoother.

• NoiseCancellationProcessor: the object that comes from the StreamVideoNoiseCancellation and performs the noise cancellation operation.
• NoiseCancellationFilter: an object that conforms to StreamVideo.AudioFilter and performs all tasks required for the noise cancellation session (e.g request start/stop).

// Firstly you initalize the processor.
let processor = NoiseCancellationProcessor()

// Secondly you instantiate the NoiseCancellationFilter. You can use any name, but it needs to be unique compared to other AudioFilters you may be using.
let noiseCancellationFilter = NoiseCancellationFilter(
    name: "noise-cancellation",
    initialize: processor.initialize,
    process: processor.process,
    release: processor.release
)

Once you are able to create a NoiseCancellationFilter you can rely on Call's API and state to toggle the filter status and also observe the features availability.

Feature availability

Call.state.settings contains the noiseCancellation configuration. The configuration contains a mode property that you look into, to determine the feature's availability:

• .available The featue has been enabled on the dashboard and it's available for the call. In this case, you are free to present any noise cancellation toggle UI in your application.

tip

Even though the feature may be enabled for your call, you should note that NoiseCancellation is a very performance-heavy process. For that reason, it's recommended to only allow the feature on devices that support Apple's neuralEngine.

You can easily check if the current device has neuralEngine support by using StreamVideo.isHardwareAccelerationAvailable on your streamVideo initialized instance.

For more info you can refer to our UI docs. about Noise Cancellation.

• .disabled The feature hasn't been enabled on the dashboard or the feature isn't available for the call. In this case, you should hide any noise cancellation toggle UI in your application.

• .autoOn Similar to .available with the difference that if possible, the StreamVideo SDK will enable the filter automatically, when the user join the call.

note

The requirements for .autoOn to work properly are:

• A VideoConfig.noiseCancellationFilter value when you initialise StreamVideo
• Device has support for Apple's neuralEngine

Activate/Deactivate the filter

The NoiseCancellationFilter is an object conforming to the AudioFilter protocol. That means, you can manage it in the same manner as any other audioFilter, as described here.)

In order to support .autoOn though, the StreamVideo SDK requires us to provide it with a NoiseCancellationFilter instance. The instance you provide, will be used whenever the noise cancellation mode is .autoOn.

You can easily provide the NoiseCancellationFilter instance every time you initialize StreamVideo, like below:

// Create the NoiseCancellationFilter like the example above.

// Then you create VideoConfig instance that includes our NoiseCancellationFilter.
let videoConfig = VideoConfig(noiseCancellationFilter: noiseCancellationFilter)

// Finally, you create the StreamVideo instance by passing in our videoConfig.
let streamVideo = StreamVideo(
    apiKey: apiKey,
    user: user,
    token: token,
    videoConfig: videoConfig,
    tokenProvider: { _ in }
)