Speech is a new framework introduced in iOS 10 which gives the app the ability of speech recognition. With the help of the framework, we can transcribe the real-time or recorded audio into text. So we can imagine a note app which translates the user’s voice to text or a Siri-like app which takes the user’s voice input as command.

Request permissions

To use speech recognition, we need to request user’s permission.

In our app’s Info.plist, add a new key NSSpeechRecognitionUsageDescription to tell the user why we need the speech recognition.

Then request permission with codes like

Recorded audio

Relatively speaking,recorded audio is easier to handle.

Here is an example project SpeechRecognition.

Real-time voice

The codes are from shu223/iOS-10-Sampler.

Set up recorder

let audioSession = AVAudioSession.sharedInstance()
try audioSession.setCategory(AVAudioSessionCategoryRecord)
try audioSession.setMode(AVAudioSessionModeMeasurement)
try audioSession.setActive(true, with: .notifyOthersOnDeactivation)

Set up speech recognizer

let speechRecognizer = SFSpeechRecognizer(locale: Locale.current)!
speechRecognizer.delegate = self

recognitionRequest = SFSpeechAudioBufferRecognitionRequest()
guard let recognitionRequest = recognitionRequest else { fatalError("Unable to created a     SFSpeechAudioBufferRecognitionRequest object") }
recognitionRequest.shouldReportPartialResults = true

recognitionTask = speechRecognizer.recognitionTask(with: recognitionRequest) { result, error in
        var isFinal = false

        if let result = result {
            self.textView.text = result.bestTranscription.formattedString
            isFinal = result.isFinal

        if error != nil || isFinal {
            inputNode.removeTap(onBus: 0)

            self.recognitionRequest = nil
            self.recognitionTask = nil


    let recordingFormat = inputNode.outputFormat(forBus: 0)
    inputNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { (buffer: AVAudioPCMBuffer, when: AVAudioTime) in

Start the recorder

    try audioEngine.start()