voice-cat/clients/windows/VoiceCat.App/Audio/WasapiAudioBackend.cs

using System.Runtime.InteropServices;
using VoiceCat.Audio;
using static VoiceCat.App.Audio.InputDeviceCapture;

namespace VoiceCat.App.Audio;

public sealed class WasapiAudioBackend : IAudioDeviceBackend
{
    public IReadOnlyList<AudioDeviceInfo> Enumerate(bool input) => InputDeviceEnumerator.List(input).Select(d => new AudioDeviceInfo(d.Id, d.Name, d.IsDefault)).ToArray();
    public IAudioCapture OpenCapture(string? deviceId, bool loopback, CapturePcmHandler handler)
    {
        var capture = new InputDeviceCapture(NormalizeDeviceId(deviceId), loopback);
        capture.PcmFrameReady += (pcm, _, channels) => handler(pcm, channels);
        if (!capture.Start()) { capture.Dispose(); throw new InvalidOperationException("WASAPI capture could not start."); }
        return new Capture(capture);
    }
    public IAudioPlayback OpenPlayback(string? deviceId = null) => new Playback(deviceId);
    // The old miniaudio ABI persisted its raw device union as hex. Its Windows member
    // is a null-terminated UTF-16 WASAPI endpoint id; preserve saved selections.
    internal static string? NormalizeDeviceId(string? id)
    {
        if (string.IsNullOrEmpty(id)) return null;
        if (id.StartsWith('{')) return id;
        try
        {
            byte[] bytes = Convert.FromHexString(id);
            string decoded = System.Text.Encoding.Unicode.GetString(bytes).Split('\0')[0];
            return decoded.StartsWith('{') ? decoded : null;
        }
        catch (FormatException) { return id; }
    }
    private sealed class Capture(InputDeviceCapture capture) : IAudioCapture { public void Dispose() => capture.Dispose(); }

    private sealed class Playback : IAudioPlayback
    {
        private readonly string? deviceId;
        private readonly PcmRing pcm = new(32768);
        private readonly ManualResetEventSlim initialized = new(false);
        private readonly Thread thread;
        private volatile bool running = true;
        private bool ready;
        private int disposed;
        internal Playback(string? deviceId)
        {
            this.deviceId = deviceId;
            thread = new Thread(Work) { IsBackground = true, Name = "VoiceCat WASAPI playback" };
            thread.Start();
            if (!initialized.Wait(5000) || !ready) { Dispose(); throw new InvalidOperationException("WASAPI playback could not start."); }
        }
        public void Write(ReadOnlySpan<short> stereoPcm) => pcm.TryWrite(stereoPcm);
        private unsafe void Work()
        {
            IMMDeviceEnumerator? enumerator = null; IMMDevice? device = null; IAudioClient? client = null; IAudioRenderClient? render = null;
            using var bufferReady = new AutoResetEvent(false);
            try
            {
                enumerator = (IMMDeviceEnumerator)Activator.CreateInstance(Type.GetTypeFromCLSID(new Guid("BCDE0395-E52F-467C-8E3D-C4579291692E"))!)!;
                int result = deviceId is null ? enumerator.GetDefaultAudioEndpoint(0, 0, out device) : enumerator.GetDevice(deviceId, out device);
                if (result < 0 || device is null) return;
                Guid iid = typeof(IAudioClient).GUID;
                if (device.Activate(ref iid, 0x17, 0, out object audio) < 0) return;
                client = (IAudioClient)audio;
                WaveFormat format = new() { Format = 1, Channels = 2, Samples = 48000, Bytes = 192000, Align = 4, Bits = 16 };
                if (client.Initialize(0, 0x00040000u | 0x80000000u | 0x08000000u, 600000, 0, (nint)(&format), 0) < 0) return;
                if (client.GetBufferSize(out uint capacity) < 0 || client.SetEventHandle(bufferReady.SafeWaitHandle.DangerousGetHandle()) < 0) return;
                iid = typeof(IAudioRenderClient).GUID;
                if (client.GetService(ref iid, out object output) < 0) return;
                render = (IAudioRenderClient)output;
                if (client.Start() < 0) return;
                ready = true; initialized.Set();
                Span<short> discard = stackalloc short[1920];
                while (running)
                {
                    bufferReady.WaitOne(20); // Scheduling wait is outside the buffer-fill cycle.
                    if (!running || client.GetCurrentPadding(out uint padding) < 0) break;
                    uint frames = capacity - Math.Min(capacity, padding);
                    if (frames == 0) continue;
                    if (render.GetBuffer(frames, out nint buffer) < 0) break;
                    var destination = new Span<short>((void*)buffer, checked((int)frames * 2));
                    // Keep queued playback bounded to ~120 ms, then fill underflow with silence.
                    while (pcm.Count > 11520) pcm.Read(discard);
                    int count = pcm.Read(destination); destination[count..].Clear();
                    if (render.ReleaseBuffer(frames, 0) < 0) break;
                }
            }
            catch { ready = false; }
            finally
            {
                initialized.Set();
                try { client?.Stop(); } catch { }
                if (render is not null) Marshal.ReleaseComObject(render);
                if (client is not null) Marshal.ReleaseComObject(client);
                if (device is not null) Marshal.ReleaseComObject(device);
                if (enumerator is not null) Marshal.ReleaseComObject(enumerator);
            }
        }
        public void Dispose()
        {
            if (Interlocked.Exchange(ref disposed, 1) != 0) return;
            running = false;
            if (!thread.Join(5000)) throw new TimeoutException("WASAPI playback did not stop.");
            initialized.Dispose();
        }
    }
    [StructLayout(LayoutKind.Sequential, Pack = 2)]
    private struct WaveFormat { internal ushort Format, Channels; internal uint Samples, Bytes; internal ushort Align, Bits, Extra; }
    [ComImport, Guid("F294ACFC-3146-4483-A7BF-ADDCA7C260E2"), InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
    private interface IAudioRenderClient
    {
        [PreserveSig] int GetBuffer(uint frames, out nint data);
        [PreserveSig] int ReleaseBuffer(uint frames, uint flags);
    }
}