This GIST is for converting Speech To Text.

SpeechToText.swift

import Foundation
import Speech

protocol SpeechToTextDelegate {
    func conversionGoingOn(_ convertedString: String)
    func conversionFinished(_ convertedString: String)
}

class SpeechToText: NSObject, SFSpeechRecognizerDelegate {
   // Creating singleton
    static let sharedIntance = SpeechToText()
    private override init() {
        super.init()
        speechRecognizer?.delegate = self
        getPermissionForSpeechRecognition()
    }
    
    
    var delegateSpeechToText: SpeechToTextDelegate?
    // This is the speech recognizer
    let speechRecognizer = SFSpeechRecognizer(locale: Locale.init(identifier: "en-US"))
    
    // This is a  Request
    var regRequest: SFSpeechAudioBufferRecognitionRequest?
    var regTask: SFSpeechRecognitionTask?
    let avEngine = AVAudioEngine()
    var speechText = String()
    var timer: Timer?
    
    // Method for stop recognition
    func stopRecognition() {
        avEngine.stop()
        regRequest?.endAudio()
        avEngine.inputNode?.removeTap(onBus: 0)
        speechText = ""
        timer?.invalidate()
        regRequest = nil
        regTask = nil
    }
    
    // Method for starting recognition
    func startRecognition() {
        //Cancel task if already running
        if regTask != nil {
            regTask?.cancel()
            regTask = nil
        }
        
        //Create and AVAudioSession for audio recording
        let avAudioSession = AVAudioSession.sharedInstance()
        do {
            try avAudioSession.setCategory(AVAudioSessionCategoryRecord)
            try avAudioSession.setMode(AVAudioSessionModeMeasurement)
            try avAudioSession.setActive(true, with: .notifyOthersOnDeactivation)
        } catch {
            print("Audio Session is not active")
        }
        
        //Check the Audio input.
        guard let inputEngineNode = avEngine.inputNode else {
            fatalError("Some Error")
        }
        
        regRequest = SFSpeechAudioBufferRecognitionRequest()
        guard let recognitionRequest = regRequest else {
            fatalError("SFSpeechAudioBufferRecognitionRequest object is not created")
        }
        recognitionRequest.shouldReportPartialResults = true
        
        //Start task of speech recognition
        regTask = speechRecognizer?.recognitionTask(with: recognitionRequest, resultHandler: { (result, error) in
           
        if result != nil {
            DispatchQueue.main.async {
                self.timer?.invalidate()
                self.timer = nil
                self.timer = Timer.scheduledTimer(timeInterval: 1, target: self, selector: #selector(SpeechToText.timerStopped), userInfo: nil, repeats: false)
                self.speechText = (result?.bestTranscription.formattedString)!
                self.delegateSpeechToText?.conversionGoingOn(self.speechText)
                }
            }
        })
        
        //Set Formation of Audio Input
        inputEngineNode.removeTap(onBus: 0)
        let recordingFormat = inputEngineNode.outputFormat(forBus: 0)
        inputEngineNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { (buffer, when) in
            self.regRequest?.append(buffer)
        }
        
        avEngine.prepare()
        
        do {
            try avEngine.start()
        } catch {
            print("some error")
        }
    }
    
    // When conversation stops
    func timerStopped() {
        delegateSpeechToText?.conversionFinished(self.speechText)
        stopRecognition()
    }
}