Port managed client audio and Windows application

dotnet/src/VoiceCat.Audio/AudioDeviceBackend.cs

@@ -0,0 +1,12 @@
+namespace VoiceCat.Audio;
+
+public sealed record AudioDeviceInfo(string Id, string Name, bool IsDefault);
+public delegate void CapturePcmHandler(ReadOnlySpan<short> pcm, int channels);
+public interface IAudioCapture : IDisposable { }
+public interface IAudioPlayback : IDisposable { void Write(ReadOnlySpan<short> stereoPcm); }
+public interface IAudioDeviceBackend
+{
+    IReadOnlyList<AudioDeviceInfo> Enumerate(bool input);
+    IAudioCapture OpenCapture(string? deviceId, bool loopback, CapturePcmHandler pcm);
+    IAudioPlayback OpenPlayback(string? deviceId = null);
+}

dotnet/src/VoiceCat.Audio/AudioEngine.cs

@@ -0,0 +1,170 @@
+using System.Diagnostics;
+using VoiceCat.Protocol;
+using Voicecat.V1;
+
+namespace VoiceCat.Audio;
+
+public sealed class AudioEngine : IDisposable
+{
+    private readonly object gate = new();
+    private readonly EncodedVoiceSender sender;
+    private readonly int[] mixed = new int[1920];
+    private readonly short[] output = new short[1920];
+    private Routes routes = new([], [], 0);
+    private readonly List<(IDisposable Stream, long Epoch)> retired = [];
+    private long completedEpoch;
+    private readonly CancellationTokenSource stop = new();
+    private readonly Task maintenance;
+    private Thread? worker;
+    private int disposed;
+    public uint SampleClock { get; private set; }
+    public event MixedPcmHandler? MixedPcm;
+    public event PcmStreamHandler? StreamPcm;
+    public volatile float InputGain = 1, OutputGain = 1, VadThreshold = 0.02f;
+    public volatile bool InputNoiseReduction, MicMuted, Deafened, PushToTalk;
+    public volatile AudioInputMode InputMode = AudioInputMode.VoiceActivation;
+    public Exception? Failure { get; private set; }
+
+    public AudioEngine(EncodedVoiceSender sender, bool startWorker = true)
+    {
+        this.sender = sender;
+        maintenance = MaintainAsync();
+        if (startWorker)
+        {
+            worker = new Thread(Work) { IsBackground = true, Name = "VoiceCat managed audio" };
+            worker.Start();
+        }
+    }
+
+    public void AddLocalStream(StreamInfo info, int captureChannels = 1)
+    {
+        if (captureChannels is not (1 or 2)) throw new ArgumentOutOfRangeException(nameof(captureChannels));
+        lock (gate)
+        {
+            ObjectDisposedException.ThrowIf(disposed != 0, this);
+            var stream = new LocalStream(info, captureChannels);
+            Routes previous = routes;
+            var locals = previous.Local.Where(s => s.Info.StreamId != info.StreamId).Append(stream).ToArray();
+            Publish(locals, previous.Remote);
+        }
+    }
+    public void RemoveLocalStream(uint streamId)
+    {
+        lock (gate) Publish(routes.Local.Where(s => s.Info.StreamId != streamId).ToArray(), routes.Remote);
+    }
+    public void SetCaptureChannels(uint streamId, int channels)
+    {
+        if (channels is not (1 or 2)) throw new ArgumentOutOfRangeException(nameof(channels));
+        lock (gate)
+        {
+            var current = routes.Local.FirstOrDefault(s => s.Info.StreamId == streamId) ?? throw new ArgumentException("Stream not found.");
+            if (current.CaptureChannels != channels) Publish(routes.Local.Select(s => s == current ? new LocalStream(s.Info, channels) : s).ToArray(), routes.Remote);
+        }
+    }
+
+    public void SetRemoteStreams(IReadOnlyList<User> users, uint selfId, uint channelId)
+    {
+        lock (gate)
+        {
+            var next = new List<ReceiveStream>();
+            foreach (User user in users.Where(u => u.Id != selfId && u.ChannelId == channelId))
+                foreach (StreamInfo info in user.Streams)
+                {
+                    var previous = routes.Remote.FirstOrDefault(s => s.UserId == user.Id && s.Info.Equals(info));
+                    next.Add(previous ?? new ReceiveStream(user.Id, info));
+                }
+            Publish(routes.Local, next.ToArray());
+        }
+    }
+
+    public bool FeedPcm(uint streamId, ReadOnlySpan<short> pcm, int channels)
+    {
+        foreach (LocalStream stream in Volatile.Read(ref routes).Local)
+            if (stream.Info.StreamId == streamId) return stream.Feed(pcm, channels);
+        return false;
+    }
+    public void Receive(VoiceFrameHeader header, ReadOnlySpan<byte> packet)
+    {
+        foreach (ReceiveStream stream in Volatile.Read(ref routes).Remote) if (stream.Info.Ssrc == header.Ssrc) { stream.Enqueue(header, packet); return; }
+    }
+    public (float Level, bool Talking) GetLocalLevel(uint streamId)
+    {
+        foreach (LocalStream stream in Volatile.Read(ref routes).Local) if (stream.Info.StreamId == streamId) return (stream.Level, stream.Talking);
+        return default;
+    }
+    public void SetRemotePlayback(uint userId, uint streamId, float gain, bool muted, bool noiseReduction)
+    {
+        if (!float.IsFinite(gain) || gain < 0 || gain > 4) throw new ArgumentOutOfRangeException(nameof(gain));
+        foreach (var stream in Volatile.Read(ref routes).Remote)
+            if (stream.UserId == userId && stream.Info.StreamId == streamId) { stream.Gain = gain; stream.Muted = muted; stream.NoiseReduction = noiseReduction; return; }
+    }
+    public (float Gain, bool Muted, bool NoiseReduction)? GetRemotePlayback(uint userId, uint streamId)
+    {
+        foreach (var stream in Volatile.Read(ref routes).Remote) if (stream.UserId == userId && stream.Info.StreamId == streamId) return (stream.Gain, stream.Muted, stream.NoiseReduction);
+        return null;
+    }
+
+    private void Publish(LocalStream[] local, ReceiveStream[] remote)
+    {
+        Routes previous = routes; var next = new Routes(local, remote, previous.Epoch + 1);
+        foreach (var stream in previous.Local) if (!local.Contains(stream)) retired.Add((stream, next.Epoch));
+        foreach (var stream in previous.Remote) if (!remote.Contains(stream)) retired.Add((stream, next.Epoch));
+        Volatile.Write(ref routes, next);
+    }
+
+    // One audio owner calls this. No allocation, waiting, lock, registry mutation or disposal
+    // occurs inside a mix cycle. Callbacks receive borrowed spans and must follow that rule.
+    internal void ProcessCycle()
+    {
+        Routes current = Volatile.Read(ref routes);
+        mixed.AsSpan().Clear();
+        foreach (var stream in current.Local) stream.Process(this, sender);
+        foreach (var stream in current.Remote) stream.Mix(mixed, Deafened, StreamPcm);
+        float gain = Deafened ? 0 : OutputGain;
+        for (int i = 0; i < output.Length; i++) output[i] = (short)Math.Clamp((int)(mixed[i] * gain), short.MinValue, short.MaxValue);
+        MixedPcm?.Invoke(output);
+        SampleClock = unchecked(SampleClock + 960);
+        Volatile.Write(ref completedEpoch, current.Epoch);
+    }
+
+    private void Work()
+    {
+        long deadline = Stopwatch.GetTimestamp();
+        try
+        {
+            while (!stop.IsCancellationRequested)
+            {
+                ProcessCycle();
+                deadline += Stopwatch.Frequency / 50;
+                double remaining = (deadline - Stopwatch.GetTimestamp()) * 1000.0 / Stopwatch.Frequency;
+                if (remaining > 0) Thread.Sleep((int)Math.Ceiling(remaining));
+                else if (remaining < -100) deadline = Stopwatch.GetTimestamp();
+            }
+        }
+        catch (Exception exception) { Failure = exception; stop.Cancel(); }
+    }
+    private async Task MaintainAsync()
+    {
+        try
+        {
+            using var timer = new PeriodicTimer(TimeSpan.FromMilliseconds(20));
+            while (await timer.WaitForNextTickAsync(stop.Token).ConfigureAwait(false))
+                lock (gate)
+                    for (int i = retired.Count - 1; i >= 0; i--)
+                        if (retired[i].Epoch <= Volatile.Read(ref completedEpoch)) { retired[i].Stream.Dispose(); retired.RemoveAt(i); }
+        }
+        catch (OperationCanceledException) when (stop.IsCancellationRequested) { }
+    }
+    public void Dispose()
+    {
+        if (Interlocked.Exchange(ref disposed, 1) != 0) return;
+        stop.Cancel(); worker?.Join(); maintenance.GetAwaiter().GetResult();
+        lock (gate)
+        {
+            foreach (var stream in routes.Local) stream.Dispose(); foreach (var stream in routes.Remote) stream.Dispose();
+            foreach (var stream in retired) stream.Stream.Dispose(); retired.Clear();
+            routes = new([], [], routes.Epoch + 1);
+        }
+    }
+    private sealed record Routes(LocalStream[] Local, ReceiveStream[] Remote, long Epoch);
+}

dotnet/src/VoiceCat.Audio/LocalStream.cs

@@ -0,0 +1,115 @@
+using VoiceCat.Codec;
+using VoiceCat.Dsp;
+using VoiceCat.Protocol;
+using Voicecat.V1;
+using OpusApplication = Voicecat.V1.OpusApplication;
+
+namespace VoiceCat.Audio;
+
+public enum AudioInputMode { VoiceActivation, PushToTalk, AlwaysOn }
+public delegate bool EncodedVoiceSender(uint ssrc, uint timestamp, ReadOnlySpan<byte> payload, VoiceFrameFlags flags);
+public delegate void PcmStreamHandler(uint userId, uint streamId, ReadOnlySpan<short> pcm, int channels);
+public delegate void MixedPcmHandler(ReadOnlySpan<short> stereoPcm);
+
+internal sealed class LocalStream : IDisposable
+{
+    internal readonly StreamInfo Info;
+    internal readonly int CaptureChannels;
+    internal readonly PcmRing Input = new(16384);
+    internal volatile float Level;
+    internal volatile bool Talking;
+    private readonly OpusEncoder encoder;
+    private readonly RnnoiseProcessor left, right;
+    private readonly EnergyVadProcessor vad = new();
+    private readonly short[] capture = new short[1920], wire = new short[5760 * 2], mono = new short[960];
+    private readonly byte[] packet = new byte[1275];
+    private readonly short[] converted = new short[16384];
+    private int feeding;
+    private int buffered;
+    private uint timestamp;
+    private bool wasTransmitting, marker;
+
+    internal bool Feed(ReadOnlySpan<short> pcm, int channels)
+    {
+        if (channels is not (1 or 2) || pcm.Length % channels != 0 || pcm.Length / channels * CaptureChannels > converted.Length || Interlocked.CompareExchange(ref feeding, 1, 0) != 0) return false;
+        try
+        {
+            if (channels == CaptureChannels) return Input.TryWrite(pcm);
+            int frames = pcm.Length / channels;
+            for (int i = 0; i < frames; i++)
+                if (CaptureChannels == 1) converted[i] = (short)(((int)pcm[i * 2] + pcm[i * 2 + 1]) / 2);
+                else { converted[2 * i] = pcm[i]; converted[2 * i + 1] = pcm[i]; }
+            return Input.TryWrite(converted.AsSpan(0, frames * CaptureChannels));
+        }
+        finally { Volatile.Write(ref feeding, 0); }
+    }
+
+    internal LocalStream(StreamInfo stream, int captureChannels)
+    {
+        Info = stream.Clone(); CaptureChannels = captureChannels;
+        encoder = new(new()
+        {
+            Channels = stream.Audio.Mode == ChannelMode.ModeStereo ? 2 : 1,
+            FrameDurationMilliseconds = checked((int)stream.Audio.FrameMs), MaximumBandwidthHz = checked((int)stream.Audio.SampleRate),
+            Bitrate = checked((int)stream.Audio.BitrateBps), Complexity = checked((int)stream.Audio.Complexity),
+            ExpectedPacketLossPercent = checked((int)stream.Audio.ExpectedPacketLoss), ForwardErrorCorrection = stream.Audio.Fec,
+            DiscontinuousTransmission = stream.Audio.Dtx, DeepRedundancy = stream.Audio.Dred,
+            Application = stream.Audio.Application switch { OpusApplication.OpusAudio => VoiceCat.Codec.OpusApplication.Audio, OpusApplication.OpusLowdelay => VoiceCat.Codec.OpusApplication.LowDelay, _ => VoiceCat.Codec.OpusApplication.Voip }
+        });
+        try { left = new(); } catch { encoder.Dispose(); throw; }
+        try { right = new(); } catch { left.Dispose(); encoder.Dispose(); throw; }
+    }
+
+    internal void Process(AudioEngine engine, EncodedVoiceSender sender)
+    {
+        var input = capture.AsSpan(0, 960 * CaptureChannels);
+        if (Input.Count < input.Length) { Level = 0; Talking = false; buffered = 0; wasTransmitting = false; return; }
+        while (Input.Count > input.Length * 6) Input.Read(input);
+        if (buffered == 0) timestamp = engine.SampleClock;
+        Input.Read(input);
+        bool mic = Info.Kind == StreamKind.StreamMic;
+        if (mic && engine.InputNoiseReduction)
+        {
+            if (CaptureChannels == 1) left.Process(input);
+            else
+            {
+                for (int i = 0; i < 960; i++) mono[i] = input[2 * i]; left.Process(mono);
+                for (int i = 0; i < 960; i++) input[2 * i] = mono[i];
+                for (int i = 0; i < 960; i++) mono[i] = input[2 * i + 1]; right.Process(mono);
+                for (int i = 0; i < 960; i++) input[2 * i + 1] = mono[i];
+            }
+        }
+        float gain = mic ? engine.InputGain : 1;
+        double energy = 0;
+        for (int i = 0; i < input.Length; i++) { input[i] = (short)Math.Clamp((int)(input[i] * gain), short.MinValue, short.MaxValue); energy += (double)input[i] * input[i]; }
+        Level = (float)(Math.Sqrt(energy / input.Length) / 32768);
+        vad.Threshold = engine.VadThreshold;
+        bool transmit = !mic || !engine.MicMuted && engine.InputMode switch
+        {
+            AudioInputMode.AlwaysOn => true, AudioInputMode.PushToTalk => engine.PushToTalk,
+            _ => vad.Process(input)
+        };
+        Talking = transmit && Level > 0.001f;
+        if (!transmit) { buffered = 0; wasTransmitting = false; return; }
+        if (!wasTransmitting) marker = true;
+        wasTransmitting = true;
+        int channels = encoder.Options.Channels;
+        for (int i = 0; i < 960; i++)
+        {
+            if (channels == 1) wire[buffered + i] = CaptureChannels == 1 ? input[i] : (short)(((int)input[2 * i] + input[2 * i + 1]) / 2);
+            else { wire[buffered + 2 * i] = input[i * CaptureChannels]; wire[buffered + 2 * i + 1] = input[i * CaptureChannels + CaptureChannels - 1]; }
+        }
+        buffered += 960 * channels;
+        int frame = encoder.Options.SamplesPerChannel * channels;
+        while (buffered >= frame)
+        {
+            int length = encoder.Encode(wire.AsSpan(0, frame), packet);
+            VoiceFrameFlags flags = (Info.Audio.Fec ? VoiceFrameFlags.FecPresent : VoiceFrameFlags.None) | (marker ? VoiceFrameFlags.Marker : VoiceFrameFlags.None);
+            sender(Info.Ssrc, timestamp, packet.AsSpan(0, length), flags); marker = false;
+            timestamp = unchecked(timestamp + (uint)encoder.Options.SamplesPerChannel);
+            buffered -= frame;
+            wire.AsSpan(frame, buffered).CopyTo(wire);
+        }
+    }
+    public void Dispose() { encoder.Dispose(); left.Dispose(); right.Dispose(); }
+}

dotnet/src/VoiceCat.Audio/PcmRing.cs

@@ -0,0 +1,34 @@
+namespace VoiceCat.Audio;
+
+// Single consumer, non-waiting producer gate. Whole writes either fit or drop, so
+// channels remain aligned and a capture thread never waits for a mixer/network owner.
+public sealed class PcmRing
+{
+    private readonly short[] samples;
+    private readonly int mask;
+    private int read, written, producer;
+    public PcmRing(int capacity = 32768)
+    {
+        if (capacity < 2 || (capacity & (capacity - 1)) != 0) throw new ArgumentOutOfRangeException(nameof(capacity));
+        samples = new short[capacity]; mask = capacity - 1;
+    }
+    public int Count => unchecked(Volatile.Read(ref written) - Volatile.Read(ref read));
+    public bool TryWrite(ReadOnlySpan<short> source)
+    {
+        if (Interlocked.CompareExchange(ref producer, 1, 0) != 0) return false;
+        try
+        {
+            int index = written;
+            if (source.Length > samples.Length - unchecked(index - Volatile.Read(ref read))) return false;
+            for (int i = 0; i < source.Length; i++) samples[(index + i) & mask] = source[i];
+            Volatile.Write(ref written, unchecked(index + source.Length)); return true;
+        }
+        finally { Volatile.Write(ref producer, 0); }
+    }
+    public int Read(Span<short> destination)
+    {
+        int index = read, count = Math.Min(destination.Length, unchecked(Volatile.Read(ref written) - index));
+        for (int i = 0; i < count; i++) destination[i] = samples[(index + i) & mask];
+        Volatile.Write(ref read, unchecked(index + count)); return count;
+    }
+}

dotnet/src/VoiceCat.Audio/ReceiveStream.cs

@@ -0,0 +1,163 @@
+using VoiceCat.Codec;
+using VoiceCat.Dsp;
+using VoiceCat.Protocol;
+using Voicecat.V1;
+
+namespace VoiceCat.Audio;
+
+internal sealed class ReceiveStream : IDisposable
+{
+    internal readonly uint UserId;
+    internal readonly StreamInfo Info;
+    internal volatile float Gain = 1;
+    internal volatile bool Muted, NoiseReduction;
+    private readonly OpusDecoder decoder;
+    private readonly OpusDeepRedundancy? dred;
+    private readonly RnnoiseProcessor left, right;
+    private readonly short[] pcm = new short[5760 * 2];
+    private readonly short[] mono = new short[960];
+    private readonly short[] block = new short[1920];
+    private readonly byte[][] packets = Enumerable.Range(0, 64).Select(_ => new byte[1275]).ToArray();
+    private readonly VoiceFrameHeader[] headers = new VoiceFrameHeader[64];
+    private readonly int[] lengths = new int[64];
+    private int read, written;
+    private readonly byte[][] jitter = Enumerable.Range(0, 6).Select(_ => new byte[1275]).ToArray();
+    private readonly uint[] timestamps = new uint[6];
+    private readonly int[] sizes = new int[6];
+    private int count, available, offset, missing, waiting;
+    private uint expected;
+    private bool started, hasTimestamp;
+    private bool hasMarker;
+    private uint lastMarker;
+    private readonly int channels, frameSamples, maximumDepth;
+    internal int Depth => count;
+    internal int ConcealedFrames { get; private set; }
+    internal int DredFrames { get; private set; }
+    internal int FecFrames { get; private set; }
+
+    internal ReceiveStream(uint userId, StreamInfo info)
+    {
+        if (info.Audio.FrameMs is not (5 or 10 or 20 or 40 or 60) || !Enum.IsDefined(info.Audio.Mode)) throw new ArgumentException("Unsupported remote audio configuration.", nameof(info));
+        UserId = userId; Info = info.Clone();
+        channels = info.Audio.Mode == ChannelMode.ModeStereo ? 2 : 1;
+        frameSamples = checked((int)info.Audio.FrameMs * 48);
+        maximumDepth = Math.Clamp(120 / (int)info.Audio.FrameMs, 2, 6);
+        decoder = new(48000, channels);
+        try { left = new(); } catch { decoder.Dispose(); throw; }
+        try { right = new(); } catch { left.Dispose(); decoder.Dispose(); throw; }
+        try { if (info.Audio.Dred) dred = new(); } catch { right.Dispose(); left.Dispose(); decoder.Dispose(); throw; }
+    }
+
+    // Only the network receive owner calls this; mixer alone consumes.
+    internal bool Enqueue(VoiceFrameHeader header, ReadOnlySpan<byte> payload)
+    {
+        int index = written;
+        if (payload.Length is < 1 or > 1275 || unchecked(index - Volatile.Read(ref read)) >= 64) return false;
+        int slot = index & 63; payload.CopyTo(packets[slot]); headers[slot] = header; lengths[slot] = payload.Length;
+        Volatile.Write(ref written, unchecked(index + 1)); return true;
+    }
+
+    private void Drain()
+    {
+        while (read != Volatile.Read(ref written))
+        {
+            int source = read & 63; uint timestamp = headers[source].Timestamp;
+            if (!hasTimestamp) { hasTimestamp = true; expected = timestamp; }
+            int delta = unchecked((int)(timestamp - expected));
+            if ((headers[source].Flags & VoiceFrameFlags.Marker) != 0 && (!hasMarker || unchecked((int)(timestamp - lastMarker)) > 0))
+            {
+                hasMarker = true; lastMarker = timestamp;
+                sizes.AsSpan().Clear(); count = available = offset = missing = waiting = 0;
+                expected = timestamp; started = false; delta = 0;
+            }
+            bool duplicate = false;
+            for (int i = 0; i < sizes.Length; i++) if (sizes[i] != 0 && timestamps[i] == timestamp) duplicate = true;
+            if ((!started || delta >= 0) && delta % frameSamples == 0 && !duplicate)
+            {
+                if (count >= maximumDepth)
+                {
+                    int oldest = Oldest(); sizes[oldest] = 0; count--;
+                }
+                int target = Array.IndexOf(sizes, 0);
+                timestamps[target] = timestamp; sizes[target] = lengths[source]; packets[source].AsSpan(0, lengths[source]).CopyTo(jitter[target]); count++;
+                int oldestRemaining = Oldest();
+                if (count >= maximumDepth && unchecked((int)(timestamps[oldestRemaining] - expected)) > 0) expected = timestamps[oldestRemaining];
+            }
+            Volatile.Write(ref read, unchecked(read + 1));
+        }
+    }
+
+    private int Oldest()
+    {
+        int oldest = -1;
+        for (int i = 0; i < sizes.Length; i++) if (sizes[i] != 0 && (oldest < 0 || unchecked((int)(timestamps[i] - timestamps[oldest])) < 0)) oldest = i;
+        return oldest;
+    }
+
+    private void Decode()
+    {
+        available = frameSamples; offset = 0;
+        int found = -1;
+        for (int i = 0; i < sizes.Length; i++) if (sizes[i] != 0 && timestamps[i] == expected) { found = i; break; }
+        bool decoded = false;
+        if (found >= 0)
+        {
+            decoded = decoder.TryDecode(jitter[found].AsSpan(0, sizes[found]), pcm, frameSamples, out int result) && result == frameSamples;
+            sizes[found] = 0; count--; missing = decoded ? 0 : missing + 1;
+        }
+        else
+        {
+            int next = Oldest(); missing++;
+            if (next >= 0 && unchecked((int)(timestamps[next] - expected)) == frameSamples)
+            {
+                var packet = jitter[next].AsSpan(0, sizes[next]);
+                if (dred?.TryRecover(decoder, packet, pcm, frameSamples) == true) { decoded = true; DredFrames++; }
+                else if (Info.Audio.Fec && decoder.TryDecode(packet, pcm, frameSamples, out int recovered, true) && recovered == frameSamples) { decoded = true; FecFrames++; }
+            }
+        }
+        int maximumConcealment = Math.Max(1, 200 / (int)Info.Audio.FrameMs);
+        if (!decoded && missing <= maximumConcealment)
+        {
+            decoded = decoder.TryDecode([], pcm, frameSamples, out _); ConcealedFrames++;
+        }
+        if (!decoded || missing > maximumConcealment) pcm.AsSpan(0, frameSamples * channels).Clear();
+        expected = unchecked(expected + (uint)frameSamples);
+    }
+
+    internal void Mix(Span<int> output, bool deafened, PcmStreamHandler? sink)
+    {
+        Drain();
+        if (!started)
+        {
+            waiting++;
+            if (!hasTimestamp || count < Math.Min(3, maximumDepth) && waiting < 3) return;
+            expected = timestamps[Oldest()]; started = true;
+        }
+        int copied = 0;
+        while (copied < 960)
+        {
+            if (available == 0) Decode();
+            int take = Math.Min(960 - copied, available);
+            var decoded = pcm.AsSpan(offset * channels, take * channels);
+            decoded.CopyTo(block.AsSpan(copied * channels));
+            available -= take; offset += take; copied += take;
+        }
+        var samples = block.AsSpan(0, 960 * channels);
+        if (NoiseReduction && Info.Kind == StreamKind.StreamMic)
+        {
+            if (channels == 1) left.Process(samples);
+            else
+            {
+                for (int i = 0; i < 960; i++) mono[i] = samples[2 * i]; left.Process(mono);
+                for (int i = 0; i < 960; i++) samples[2 * i] = mono[i];
+                for (int i = 0; i < 960; i++) mono[i] = samples[2 * i + 1]; right.Process(mono);
+                for (int i = 0; i < 960; i++) samples[2 * i + 1] = mono[i];
+            }
+        }
+        float gain = Muted || deafened ? 0 : Gain;
+        for (int i = 0; i < samples.Length; i++) samples[i] = (short)Math.Clamp((int)(samples[i] * gain), short.MinValue, short.MaxValue);
+        sink?.Invoke(UserId, Info.StreamId, samples, channels);
+        for (int i = 0; i < 960; i++) { output[i * 2] += samples[i * channels]; output[i * 2 + 1] += samples[i * channels + channels - 1]; }
+    }
+    public void Dispose() { decoder.Dispose(); dred?.Dispose(); left.Dispose(); right.Dispose(); }
+}

dotnet/src/VoiceCat.Audio/VoiceCat.Audio.csproj

@@ -0,0 +1,8 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <ItemGroup>
+    <ProjectReference Include="../VoiceCat.Codec/VoiceCat.Codec.csproj" />
+    <ProjectReference Include="../VoiceCat.Dsp/VoiceCat.Dsp.csproj" />
+    <ProjectReference Include="../VoiceCat.Protocol/VoiceCat.Protocol.csproj" />
+    <InternalsVisibleTo Include="VoiceCat.Tests" />
+  </ItemGroup>
+</Project>

dotnet/src/VoiceCat.Audio/packages.lock.json

@@ -0,0 +1,24 @@
+{
+  "version": 1,
+  "dependencies": {
+    "net10.0": {
+      "Google.Protobuf": {
+        "type": "Transitive",
+        "resolved": "3.36.1",
+        "contentHash": "77AqPEoaY1ODE+syYBHti0jXiwQq0J/fUr/fRyYhNlc9oKtH5dZZEr/OLKtdKNVG83PRnCYB2r8B80ZrObzOGQ=="
+      },
+      "voicecat.codec": {
+        "type": "Project"
+      },
+      "voicecat.dsp": {
+        "type": "Project"
+      },
+      "voicecat.protocol": {
+        "type": "Project",
+        "dependencies": {
+          "Google.Protobuf": "[3.36.1, )"
+        }
+      }
+    }
+  }
+}

dotnet/src/VoiceCat.Audio/packages.publish.win-x64.lock.json

@@ -0,0 +1,31 @@
+{
+  "version": 1,
+  "dependencies": {
+    "net10.0": {
+      "Microsoft.NET.ILLink.Tasks": {
+        "type": "Direct",
+        "requested": "[10.0.7, )",
+        "resolved": "10.0.7",
+        "contentHash": "AA/yhzFHNtQZXLdqjzujPy25G8EWwGWsAnxOE2zYSBoT/8QHP6ketN3CToD3DFreO653ipUwnKHo22B8AlBMCw=="
+      },
+      "Google.Protobuf": {
+        "type": "Transitive",
+        "resolved": "3.36.1",
+        "contentHash": "77AqPEoaY1ODE+syYBHti0jXiwQq0J/fUr/fRyYhNlc9oKtH5dZZEr/OLKtdKNVG83PRnCYB2r8B80ZrObzOGQ=="
+      },
+      "voicecat.codec": {
+        "type": "Project"
+      },
+      "voicecat.dsp": {
+        "type": "Project"
+      },
+      "voicecat.protocol": {
+        "type": "Project",
+        "dependencies": {
+          "Google.Protobuf": "[3.36.1, )"
+        }
+      }
+    },
+    "net10.0/win-x64": {}
+  }
+}