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

using VoiceCat.Audio;
using VoiceCat.Interop;

namespace VoiceCat.App.Audio;

// Capture threads own their ring producers; the mix thread alone consumes them.
public sealed class ProcessAudioMixer : IDisposable
{
    private sealed class Input
    {
        internal readonly PcmRing Ring = new(16384);
        private readonly short[] stereo = new short[1920];
        internal void Feed(short[] pcm, int channels)
        {
            if (channels == 2) { Ring.TryWrite(pcm); return; }
            if (channels != 1 || pcm.Length > 960) return;
            for (int i = 0; i < pcm.Length; i++) stereo[2 * i] = stereo[2 * i + 1] = pcm[i];
            Ring.TryWrite(stereo.AsSpan(0, pcm.Length * 2));
        }
    }
    private readonly List<ProcessLoopbackCapture> captures = [];
    private Input[] inputs = [];
    private Thread? thread;
    private volatile bool running;
    private VoiceCatClient? client;
    private uint streamId;
    public void Start(AppAudioScope scope, VoiceCatClient client, uint streamId)
    {
        if (running) return;
        this.client = client; this.streamId = streamId;
        var specs = ResolveCaptures(scope);
        inputs = specs.Select(_ => new Input()).ToArray();
        try
        {
            for (int i = 0; i < specs.Count; i++)
            {
                Input input = inputs[i]; var (pid, mode) = specs[i];
                var capture = new ProcessLoopbackCapture(pid, mode);
                capture.PcmFrameReady += (pcm, _, channels) => input.Feed(pcm, channels);
                captures.Add(capture);
                if (!capture.Start()) throw new InvalidOperationException("Process audio capture could not start.");
            }
            if (inputs.Length == 0) return;
            running = true;
            thread = new Thread(MixLoop) { IsBackground = true, Name = "ProcessAudioMixer" }; thread.Start();
        }
        catch { Stop(); throw; }
    }
    public void Stop()
    {
        running = false;
        if (thread is not null && !thread.Join(5000)) throw new TimeoutException("Process audio mixer did not stop.");
        foreach (var capture in captures) capture.Dispose();
        captures.Clear(); inputs = []; thread = null;
    }
    public void Dispose() => Stop();
    private void MixLoop()
    {
        var frame = new short[1920]; var mix = new short[1920]; var sums = new int[1920];
        long deadline = System.Diagnostics.Stopwatch.GetTimestamp();
        while (running)
        {
            sums.AsSpan().Clear();
            foreach (Input input in inputs)
            {
                while (input.Ring.Count > 11520) input.Ring.Read(frame);
                frame.AsSpan().Clear(); input.Ring.Read(frame);
                for (int i = 0; i < frame.Length; i++) sums[i] += frame[i];
            }
            for (int i = 0; i < mix.Length; i++) mix[i] = (short)Math.Clamp(sums[i], short.MinValue, short.MaxValue);
            client?.StreamFeedPcm(streamId, mix, 960, 2);
            deadline += System.Diagnostics.Stopwatch.Frequency / 50;
            double wait = (deadline - System.Diagnostics.Stopwatch.GetTimestamp()) * 1000.0 / System.Diagnostics.Stopwatch.Frequency;
            if (wait > 0) Thread.Sleep((int)Math.Ceiling(wait));
            else if (wait < -100) deadline = System.Diagnostics.Stopwatch.GetTimestamp();
        }
    }
    private static List<(int pid, ProcessLoopbackCapture.Mode mode)> ResolveCaptures(AppAudioScope scope) => scope switch
    {
        OnlyApps o => o.Pids.Select(p => (p, ProcessLoopbackCapture.Mode.Include)).ToList(),
        AllExceptApps a when a.Pids.Count > 0 => [(a.Pids[0], ProcessLoopbackCapture.Mode.Exclude)],
        EntireDesktop { ExcludeSelf: true } => [(Environment.ProcessId, ProcessLoopbackCapture.Mode.Exclude)],
        _ => [],
    };
}