Add managed macOS Core Audio voice path

2026-09-16 22:05:40 +02:00
parent e5ff484cac
commit edd7783a5c
8 changed files with 314 additions and 34 deletions
@@ -0,0 +1,200 @@
+using System.Runtime.InteropServices;
+using AVFoundation;
+using Foundation;
+using VoiceCat.Audio;
+
+namespace VoiceCat.Mac;
+
+internal sealed class MacAudioBackend : IAudioDeviceBackend
+{
+    private static readonly AudioDeviceInfo[] DefaultInput = [new("", "System default microphone", true)];
+    private static readonly AudioDeviceInfo[] DefaultOutput = [new("", "System default output", true)];
+    private Exception? failure;
+
+    internal Exception? Failure => Volatile.Read(ref failure);
+
+    public IReadOnlyList<AudioDeviceInfo> Enumerate(bool input) => input ? DefaultInput : DefaultOutput;
+
+    public IAudioCapture OpenCapture(string? deviceId, bool loopback, CapturePcmHandler pcm)
+    {
+        if (loopback) throw new NotSupportedException("Screen audio uses ScreenCaptureKit, not microphone capture.");
+        if (!string.IsNullOrEmpty(deviceId)) throw new NotSupportedException("Selecting a non-default Core Audio input is not implemented yet.");
+        return new Capture(pcm, ReportFailure);
+    }
+
+    public IAudioPlayback OpenPlayback(string? deviceId = null)
+    {
+        if (!string.IsNullOrEmpty(deviceId)) throw new NotSupportedException("Selecting a non-default Core Audio output is not implemented yet.");
+        return new Playback(ReportFailure);
+    }
+
+    private void ReportFailure(Exception exception) => Interlocked.CompareExchange(ref failure, exception, null);
+
+    private sealed class Capture : IAudioCapture
+    {
+        private readonly CapturePcmHandler handler;
+        private readonly AVAudioEngine engine = new();
+        private readonly AVAudioFormat outputFormat;
+        private readonly AVAudioConverter converter;
+        private readonly AVAudioPcmBuffer converted;
+        private readonly AVAudioConverterInputHandler inputProvider;
+        private readonly Action<Exception> reportFailure;
+        private AVAudioPcmBuffer? pendingInput;
+        private bool inputProvided;
+        private int disposed;
+
+        internal Capture(CapturePcmHandler handler, Action<Exception> reportFailure)
+        {
+            this.handler = handler;
+            this.reportFailure = reportFailure;
+            AVAudioInputNode input = engine.InputNode;
+            AVAudioFormat inputFormat = input.GetBusOutputFormat(0);
+            uint channels = Math.Clamp(inputFormat.ChannelCount, 1u, 2u);
+            outputFormat = new(AVAudioCommonFormat.PCMInt16, 48_000, channels, true);
+            converter = new(inputFormat, outputFormat);
+            uint outputCapacity = checked((uint)Math.Ceiling(4_096 * 48_000 / inputFormat.SampleRate) + 64);
+            converted = new(outputFormat, outputCapacity);
+            inputProvider = ProvideInput;
+            input.InstallTapOnBus(0, 960, inputFormat, Convert);
+            engine.Prepare();
+            if (!engine.StartAndReturnError(out var error))
+            {
+                input.RemoveTapOnBus(0);
+                converted.Dispose();
+                converter.Dispose();
+                outputFormat.Dispose();
+                engine.Dispose();
+                throw new InvalidOperationException("Core Audio capture could not start: " + error.LocalizedDescription);
+            }
+        }
+
+        private unsafe void Convert(AVAudioPcmBuffer input, AVAudioTime _time)
+        {
+            try
+            {
+                pendingInput = input;
+                inputProvided = false;
+                converted.FrameLength = 0;
+                AVAudioConverterOutputStatus result = converter.ConvertToBuffer(converted, out NSError? error, inputProvider);
+                if (result == AVAudioConverterOutputStatus.Error)
+                {
+                    reportFailure(new InvalidOperationException("Core Audio input conversion failed: " + error?.LocalizedDescription));
+                    return;
+                }
+                if (converted.FrameLength == 0) return;
+                nint samples = Marshal.ReadIntPtr(converted.Int16ChannelData);
+                if (samples == 0) return;
+                int channels = checked((int)outputFormat.ChannelCount);
+                int remainingFrames = checked((int)converted.FrameLength);
+                var pcm = new ReadOnlySpan<short>((void*)samples, checked(remainingFrames * channels));
+                while (remainingFrames > 0)
+                {
+                    int frames = Math.Min(960, remainingFrames);
+                    handler(pcm[..(frames * channels)], channels);
+                    pcm = pcm[(frames * channels)..];
+                    remainingFrames -= frames;
+                }
+            }
+            catch (Exception exception) { reportFailure(exception); }
+            finally { pendingInput = null; }
+        }
+
+        private AVAudioBuffer ProvideInput(uint _, out AVAudioConverterInputStatus status)
+        {
+            if (!inputProvided && pendingInput is { } input)
+            {
+                inputProvided = true;
+                status = AVAudioConverterInputStatus.HaveData;
+                return input;
+            }
+            status = AVAudioConverterInputStatus.NoDataNow;
+            return null!;
+        }
+
+        public void Dispose()
+        {
+            if (Interlocked.Exchange(ref disposed, 1) != 0) return;
+            engine.InputNode.RemoveTapOnBus(0);
+            engine.Stop();
+            converted.Dispose();
+            converter.Dispose();
+            outputFormat.Dispose();
+            engine.Dispose();
+        }
+    }
+
+    private sealed class Playback : IAudioPlayback
+    {
+        private readonly PcmRing pcm = new(32_768);
+        private readonly AVAudioEngine engine = new();
+        private readonly AVAudioFormat format;
+        private readonly AVAudioSourceNode source;
+        private readonly Action<Exception> reportFailure;
+        private int callbackFailed;
+        private int disposed;
+
+        internal Playback(Action<Exception> reportFailure)
+        {
+            this.reportFailure = reportFailure;
+            format = new(AVAudioCommonFormat.PCMInt16, 48_000, 2, true);
+            source = new(format, Render);
+            engine.AttachNode(source);
+            engine.Connect(source, engine.MainMixerNode, format);
+            engine.Prepare();
+            if (!engine.StartAndReturnError(out var error))
+            {
+                engine.DetachNode(source);
+                source.Dispose();
+                format.Dispose();
+                engine.Dispose();
+                throw new InvalidOperationException("Core Audio playback could not start: " + error.LocalizedDescription);
+            }
+        }
+
+        public void Write(ReadOnlySpan<short> stereoPcm) => pcm.TryWrite(stereoPcm);
+
+        private unsafe int Render(IntPtr isSilence, IntPtr _, uint frameCount, IntPtr outputData)
+        {
+            try
+            {
+                int bufferCount = Marshal.ReadInt32(outputData);
+                int firstBuffer = IntPtr.Size == 8 ? 8 : 4;
+                if (bufferCount != 1) return Fail("Core Audio returned a non-interleaved playback buffer.");
+                int channels = Marshal.ReadInt32(outputData, firstBuffer);
+                int byteCount = Marshal.ReadInt32(outputData, firstBuffer + 4);
+                nint data = Marshal.ReadIntPtr(outputData, firstBuffer + 8);
+                int requested = checked((int)frameCount * 2);
+                if (channels != 2 || data == 0 || byteCount < requested * sizeof(short)) return Fail("Core Audio returned an invalid stereo playback buffer.");
+                var output = new Span<short>((void*)data, requested);
+                Span<short> discard = stackalloc short[1_920];
+                while (pcm.Count > 11_520) pcm.Read(discard[..Math.Min(discard.Length, pcm.Count - 11_520)]);
+                int read = pcm.Read(output);
+                output[read..].Clear();
+                if (isSilence != IntPtr.Zero) Marshal.WriteByte(isSilence, read == 0 ? (byte)1 : (byte)0);
+                return 0;
+            }
+            catch (Exception exception)
+            {
+                if (Interlocked.Exchange(ref callbackFailed, 1) == 0) reportFailure(exception);
+                return -1;
+            }
+        }
+
+        private int Fail(string message)
+        {
+            if (Interlocked.Exchange(ref callbackFailed, 1) == 0) reportFailure(new InvalidOperationException(message));
+            return -1;
+        }
+
+        public void Dispose()
+        {
+            if (Interlocked.Exchange(ref disposed, 1) != 0) return;
+            engine.Stop();
+            engine.DetachNode(source);
+            source.Dispose();
+            format.Dispose();
+            engine.Dispose();
+        }
+    }
+
+}