ios – Swift – MacOS – Audio enter timestamp overlaps or precedes prior audio enter

I get an error that I’m not in a position to see any useful resource across the web for. I attempted a number of various things however none are in a position to resolve this subject: Audio enter timestamp overlaps or precedes prior audio enter
Context: I used to be taking part in round with some code on my Mac, and I maintain getting the error after half a minute or so if operating this code which transcribes speech to textual content utilizing the Apple Speech framework’s SpeechTranscriber. I’ve added a number of feedback to assist clarify the code too. Any assist could be significantly appreciated! I get the identical subject with DictationTranscriber as effectively, which made me really feel like the problem is someplace in my inputNode.installTap() code block, however I simply can’t determine it out

non-public func startRecording() async 
{
    do {
        // 1. Request microphone entry
        let permissionGranted = await AVAudioApplication.requestRecordPermission()
        guard permissionGranted else {
            transcribedText = "Microphone entry denied."
            return
        }
        
        let targetLocale = Locale(identifier: selectedLanguage)
        
        // 2 & 3. Create the suitable transcriber kind safely
        var resolvedTranscriber: SpeechModule? = nil
        var isUsingSpeechTranscriber = false
        
        // Strive SpeechTranscriber first if supported equal asset is confirmed
        if let supportedLocale = await SpeechTranscriber.supportedLocale(equivalentTo: targetLocale) {
            do {
                // Reserve the asset to register the bundle subscription
                attempt await AssetInventory.reserve(locale: supportedLocale)
                
                let transcriberInstance = SpeechTranscriber(locale: supportedLocale, preset: .progressiveTranscription)
                
                // 4. Obtain and Set up Property (Essential test inside a sandbox do-catch)
                if let request = attempt await AssetInventory.assetInstallationRequest(supporting: [transcriberInstance]) {
                    isDownloadingAssets = true
                    transcribedText = "Downloading language belongings..."
                    attempt await request.downloadAndInstall()
                    isDownloadingAssets = false
                }
                
                resolvedTranscriber = transcriberInstance
                isUsingSpeechTranscriber = true
                print("Efficiently initialized high-end SpeechTranscriber for (supportedLocale.identifier)")
                
            } catch {
                isDownloadingAssets = false
                print("SpeechTranscriber subscription or obtain failed: (error.localizedDescription)")
                print("Falling again to DictationTranscriber for system-supported language workflow.")
            }
        }
                
        guard let transcriberModule = resolvedTranscriber else { return }
        transcribedText = "Listening..."
        var inputNode = engine.inputNode
        var recordingFormat = inputNode.inputFormat(forBus: 0)
        let (inputSequence, builder) = AsyncStream.makeStream(of: AnalyzerInput.self)
        self.inputBuilder = builder

        if let selectedInputID = selectedInputID, let audioUnit = inputNode.audioUnit {
            print("-A1")
            var deviceID = selectedInputID
            let error = AudioUnitSetProperty(
                audioUnit,
                kAudioOutputUnitProperty_CurrentDevice,
                kAudioUnitScope_Global,
                0,
                &deviceID,
                UInt32(MemoryLayout.measurement)
            )
            if error != noErr {
                print("Warning: Couldn't set enter system (Error (error))")
            }
            // Inform the engine to flush modifications and adapt to the brand new {hardware} system
            engine.reset()
            engine.put together()
            inputNode.removeTap(onBus: 0)
            inputNode = engine.inputNode //
            recordingFormat = inputNode.inputFormat(forBus: 0)
            print("Recording format pattern fee: (recordingFormat.sampleRate)Hz") // It will now accurately output 96000Hz!
        }

        // 5. Setup AsyncStream for Audio Enter

        // Faucet the microphone to get audio buffers
        let hardwareFormat = inputNode.inputFormat(forBus: 0)
        guard let targetSpeechFormat = AVAudioFormat(
            commonFormat: .pcmFormatInt16,
            sampleRate: 16000.0,            // CRITICAL: Downsample to 16kHz
            channels: 1,                    // Dictation works greatest with mono (1 channel)
            interleaved: false
        ) else {
            print("Error creating goal speech format.")
            return
        }

        // 3. Initialize the audio converter that handles each format conversion AND downsampling
        guard let audioConverter = AVAudioConverter(from: hardwareFormat, to: targetSpeechFormat) else {
            print("Couldn't create AVAudioConverter for 16kHz conversion.")
            return
        }
        print("A0")
        self.processedConvertedFrames = 0

        // 2. Replace your faucet block
        inputNode.installTap(onBus: 0, bufferSize: 4096, format: hardwareFormat) { [weak self] (incomingBuffer, when) in
            guard let self = self else { return }
            
            let sampleRateRatio = targetSpeechFormat.sampleRate / hardwareFormat.sampleRate
            let targetFrameCapacity = AVAudioFrameCount(Double(incomingBuffer.frameLength) * sampleRateRatio)
            
            guard let convertedBuffer = AVAudioPCMBuffer(pcmFormat: targetSpeechFormat, frameCapacity: targetFrameCapacity) else {
                return
            }
            
            var error: NSError? = nil
            var hasProvidedData = false // Monitor if we have already handed the converter our knowledge
            
            let standing = audioConverter.convert(to: convertedBuffer, error: &error) { inNumPackets, outStatus in
                // If the converter asks for MORE knowledge in the identical faucet cycle, we should inform it to attend.
                if hasProvidedData {
                    outStatus.pointee = .noDataNow
                    return nil
                }
                outStatus.pointee = .haveData
                hasProvidedData = true
                return incomingBuffer
            }
            
            if standing == .error || error != nil {
                print("Conversion failed: (error?.localizedDescription ?? "Unknown error")")
                return
            }
            
            // Reject empty buffers!
            // A 0-length buffer generates an similar timestamp to the earlier body, inflicting the Code 2 Crash.
            guard convertedBuffer.frameLength > 0 else {
                return
            }
            
            // Calculate exact timeline based mostly on whole processed frames
            let currentSeconds = Double(self.processedConvertedFrames) / targetSpeechFormat.sampleRate
            let cmTime = CMTime(seconds: currentSeconds, preferredTimescale: Int32(targetSpeechFormat.sampleRate))
            
            let inputSample = AnalyzerInput(buffer: convertedBuffer, bufferStartTime: cmTime)
            self.inputBuilder?.yield(inputSample)
            
            // Increment ONLY by the frames that have been efficiently created
            self.processedConvertedFrames += Int64(convertedBuffer.frameLength)
        }
        
        self.transcriptionTask = Process { @MainActor in
            print("A2")
            // Reset our textual content buffer when a brand new session begins
            self.finalizedTextHistory = ""
            var lastSegmentEndTime: TimeInterval = Date().timeIntervalSince1970
            
            do {
                if isUsingSpeechTranscriber, let speechTranscriber = transcriberModule as? SpeechTranscriber {
                    for attempt await end in speechTranscriber.outcomes {
                        let activeText = String(outcome.textual content.characters)
                        if outcome.isFinal {
                            self.finalizedTextHistory += activeText
                            self.transcribedText = self.finalizedTextHistory
                        } else {
                            self.transcribedText = self.finalizedTextHistory + activeText
                        }
                    }
                } 
            } catch {
                print("Transcription stream threw an error: (error)")
            }
        }
        print("A4")
        engine.put together()
        do {
            attempt engine.begin()
            print("Audio Engine began efficiently!")
        } catch {
            print("Error beginning audio engine: (error.localizedDescription)")
        }
        isRecording = true
        transcribedText = "" // Clear outdated textual content
//
//            // 6. Setup Analyzer with the polymorphically typed transcriberModule
        let analyzer = SpeechAnalyzer(modules: [transcriberModule])
        self.analyzer = analyzer

        analysisTask = Process {
            do {
                let lastSampleTime = attempt await analyzer.analyzeSequence(inputSequence)
                if let time = lastSampleTime {
                    attempt await analyzer.finalizeAndFinish(by: time)
                } else {
                    await analyzer.cancelAndFinishNow()
                }
            } catch {
                print("Evaluation failed: (error)")
            }
        }
        
    } catch {
        print("Failed to start out recording setup: (error)")
        stopRecording()
    }
}

That is the logs in my console after I run this, the error reveals up 30-60 seconds after it has began and works effectively for a bit:

Recording format pattern fee: 96000.0Hz
A0
A4
Audio Engine began efficiently!
A2
A3
SpeechAnalyzer: Enter loop ending with error: Error Area=SFSpeechErrorDomain Code=2 "Audio enter timestamp overlaps or precedes prior audio enter" UserInfo={NSLocalizedDescription=Audio enter timestamp overlaps or precedes prior audio enter}
Evaluation failed: Error Area=SFSpeechErrorDomain Code=2 "Audio enter timestamp overlaps or precedes prior audio enter" UserInfo={NSLocalizedDescription=Audio enter timestamp overlaps or precedes prior audio enter}
Transcription stream threw an error: Error Area=SFSpeechErrorDomain Code=2 "Audio enter timestamp overlaps or precedes prior audio enter" UserInfo={NSLocalizedDescription=Audio enter timestamp overlaps or precedes prior audio enter}