feat(clients): add aux outgoing stream (mic + second input device) on Windows + macOS

Lets a user transmit a second hardware input device (e.g. line-in / aux)
alongside the mic, with its own device picker and volume, from Audio Settings.

No core/ABI/proto changes: the aux is a VC_STREAM_AUX_DEVICE stream started
with external_feed=1 and fed via vc_stream_feed_pcm (the same external-feed
pipeline screen-audio uses). Per-kind local_streams_ already allows mic +
screen + one aux to coexist; volume is a client-side gain multiply (the core's
vc_set_input_gain is mic-only/global). Aux is always-on (core never gates
AUX_DEVICE on VAD/PTT) and is tied to the voice session.

Windows: new Audio/InputDeviceCapture.cs (WASAPI shared-mode capture from a
real input endpoint + capture-endpoint enumeration); aux section in
AudioSettingsForm.cs; lifecycle in MainForm.cs; persistence in VoiceSettings.cs.

macOS: new Audio/InputDeviceCapture.swift (AVAudioEngine input-node tap pinned
to the chosen Core Audio device + device enumeration by stable UID); aux section
in SettingsWindowController.swift; lifecycle + UserDefaults persistence in
MainWindowController.swift; file registered in project.pbxproj.

Windows verified (C# solution builds clean; aux confirmed working). macOS build
+ E2E pending a Mac.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

PROGRESS.md

@@ -10,6 +10,30 @@ up instantly. Newest status at the top.
 
 ## ▶ Where we left off / next action
 
+- **Done (2026-06-23):** **Aux outgoing stream (mic + a second input device) — Windows + macOS.**
+  Users can now transmit a second hardware input device (e.g. line-in / aux) alongside the mic, with
+  its own device picker and volume, from Audio Settings. **No core/ABI/proto changes** — the aux is
+  a `VC_STREAM_AUX_DEVICE` stream started with `external_feed=1`, captured client-side, and fed via
+  `vc_stream_feed_pcm` (the same external-feed pipeline screen-audio uses). Per-kind `local_streams_`
+  already allows mic + screen + one aux to coexist; volume is a client-side gain multiply (the core's
+  `vc_set_input_gain` is mic-only/global). The aux is always-on (the core never gates `AUX_DEVICE` on
+  VAD/PTT) and is tied to the voice session (started on Join Voice when enabled, stopped on Leave).
+  - **Windows:** new `Audio/InputDeviceCapture.cs` (WASAPI shared-mode capture from a real input
+    endpoint via `IMMDevice.Activate(IAudioClient)`, 48 kHz/s16, 20 ms frames) + `InputDeviceEnumerator`
+    (WASAPI capture-endpoint list — separate from the core's miniaudio ids). Aux section in
+    `AudioSettingsForm.cs` (enable checkbox, device combo, refresh, volume slider, accessible names,
+    live-apply + Cancel revert via callbacks). Lifecycle in `MainForm.cs` (`_auxStreamId` +
+    `InputDeviceCapture`). Persisted in `VoiceSettings.cs` (`AuxEnabled/AuxDeviceId/AuxGain`).
+  - **macOS:** new `Audio/InputDeviceCapture.swift` (AVAudioEngine input-node tap pinned to the chosen
+    Core Audio device via `kAudioOutputUnitProperty_CurrentDevice`; AVAudioConverter → 48 kHz int16;
+    20 ms framing modelled on `ScreenAudioCapture`) + `InputDeviceEnumerator` (Core Audio device list
+    by stable UID). Aux section in `SettingsWindowController.swift`; lifecycle + UserDefaults
+    persistence (`voice.aux*`) in `MainWindowController.swift`. New file added to `project.pbxproj`.
+  - **Verify status:** Windows C# solution builds clean (0 warn/0 err); `ctest` core suite unchanged
+    (no core edits). **Next (manual):** on a Mac, build `VoiceCatMac.xcodeproj`; then two-client E2E —
+    enable aux on a second input device, confirm two distinct streams for the sender and that the aux
+    volume slider moves the aux level independently of the mic; confirm persistence across relaunch.
+
 - **Done (2026-06-23):** **Input-settings persistence, mic input gain, + two iOS bugs (all 3
   clients).** Four fixes:
   1. **Input settings now persist.** Transmission mode (VAD/PTT/Always-On), VAD threshold, and the

clients/apple/macOS/VoiceCatMac.xcodeproj/project.pbxproj

@@ -31,6 +31,7 @@
 		AAAA00000000000000000048 /* UserPickerSheet.swift in Sources */ = {isa = PBXBuildFile; fileRef = AAAA00000000000000000047 /* UserPickerSheet.swift */; };
 		AAAA0000000000000000004A /* SettingsWindowController.swift in Sources */ = {isa = PBXBuildFile; fileRef = AAAA00000000000000000049 /* SettingsWindowController.swift */; };
 		AAAA0000000000000000004C /* ScreenAudioCapture.swift in Sources */ = {isa = PBXBuildFile; fileRef = AAAA0000000000000000004B /* ScreenAudioCapture.swift */; };
+		AAAA00000000000000000051 /* InputDeviceCapture.swift in Sources */ = {isa = PBXBuildFile; fileRef = AAAA00000000000000000050 /* InputDeviceCapture.swift */; };
 		AAAA0000000000000000004E /* ScreenSharePickerSheet.swift in Sources */ = {isa = PBXBuildFile; fileRef = AAAA0000000000000000004F /* ScreenSharePickerSheet.swift */; };
 /* End PBXBuildFile section */
 
@@ -45,6 +46,7 @@
 		AAAA00000000000000000019 /* ConnectWindowController.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ConnectWindowController.swift; sourceTree = "<group>"; };
 		AAAA0000000000000000001A /* MainWindowController.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = MainWindowController.swift; sourceTree = "<group>"; };
 		AAAA0000000000000000004B /* ScreenAudioCapture.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ScreenAudioCapture.swift; sourceTree = "<group>"; };
+		AAAA00000000000000000050 /* InputDeviceCapture.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = InputDeviceCapture.swift; sourceTree = "<group>"; };
 		AAAA0000000000000000001B /* AddServerSheet.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AddServerSheet.swift; sourceTree = "<group>"; };
 		AAAA0000000000000000001C /* ServerIdentitySheet.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ServerIdentitySheet.swift; sourceTree = "<group>"; };
 		AAAA0000000000000000001D /* PasswordPromptSheet.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PasswordPromptSheet.swift; sourceTree = "<group>"; };
@@ -104,6 +106,7 @@
 			isa = PBXGroup;
 			children = (
 				AAAA0000000000000000004B /* ScreenAudioCapture.swift */,
+				AAAA00000000000000000050 /* InputDeviceCapture.swift */,
 			);
 			path = Audio;
 			sourceTree = "<group>";
@@ -247,6 +250,7 @@
 				AAAA0000000000000000004E /* ScreenSharePickerSheet.swift in Sources */,
 				AAAA0000000000000000004A /* SettingsWindowController.swift in Sources */,
 				AAAA0000000000000000004C /* ScreenAudioCapture.swift in Sources */,
+				AAAA00000000000000000051 /* InputDeviceCapture.swift in Sources */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 		};

clients/apple/macOS/VoiceCatMac/Audio/InputDeviceCapture.swift

@@ -0,0 +1,229 @@
+import AVFoundation
+import CoreAudio
+
+// InputDeviceCapture — captures a single hardware INPUT device (mic / line-in / aux) on macOS and
+// emits 20 ms (960 samples/channel @ 48 kHz, interleaved int16) frames for the aux outgoing stream.
+//
+// The input-device analogue of ScreenAudioCapture (which captures system audio via
+// ScreenCaptureKit). The core already owns ONE capture device (the mic) and can't open a second
+// arbitrary input, so for the aux stream the client captures the chosen device here and feeds PCM
+// into the core via `vc_stream_feed_pcm` — the same external-feed pipeline screen audio uses.
+//
+// Device selection: an AVAudioEngine's input node wraps an AUHAL audio unit; setting
+// kAudioOutputUnitProperty_CurrentDevice on it pins capture to a specific Core Audio device. We
+// pass the device's *UID* (stable across reboots/replugs, unlike AudioDeviceID) and resolve it at
+// start. The tap delivers Float32; AVAudioConverter resamples/quantises to 48 kHz int16.
+final class InputDeviceCapture {
+
+    /// Receives a full 20 ms frame: (interleaved int16 PCM, samplesPerChannel = 960, channels).
+    typealias PcmHandler = (UnsafePointer<Int16>, Int, UInt32) -> Void
+
+    enum CaptureError: Error { case deviceNotFound, engineStartFailed(OSStatus) }
+
+    private static let frameSamplesPerChannel = 960  // 20 ms @ 48 kHz
+
+    private let deviceUID: String?  // nil = system default input device
+    private let onPcm: PcmHandler
+    private let engine = AVAudioEngine()
+    private var converter: AVAudioConverter?
+    private var outFormat: AVAudioFormat?
+    private var channels = 1
+
+    /// Interleaved int16 carry-over between tap callbacks (the tap buffer doesn't align to 20 ms),
+    /// drained in whole `frameSamplesPerChannel * channels` chunks. Only touched on the tap queue.
+    private var pending: [Int16] = []
+
+    init(deviceUID: String?, onPcm: @escaping PcmHandler) {
+        self.deviceUID = deviceUID
+        self.onPcm = onPcm
+    }
+
+    /// Begin capture. Throws if the device can't be resolved or the engine fails to start.
+    func start() throws {
+        let input = engine.inputNode
+
+        // Pin the engine's AUHAL to the chosen device (skip for default — the engine already uses
+        // the system default input). Must happen before reading inputFormat, which changes with
+        // the selected device.
+        if let deviceUID, let devId = Self.deviceID(forUID: deviceUID) {
+            var dev = devId
+            let status = AudioUnitSetProperty(input.audioUnit!,
+                                              kAudioOutputUnitProperty_CurrentDevice,
+                                              kAudioUnitScope_Global, 0,
+                                              &dev, UInt32(MemoryLayout<AudioDeviceID>.size))
+            if status != noErr { throw CaptureError.engineStartFailed(status) }
+        } else if deviceUID != nil {
+            throw CaptureError.deviceNotFound
+        }
+
+        let inFormat = input.inputFormat(forBus: 0)
+        channels = max(1, min(2, Int(inFormat.channelCount)))
+        guard let out = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: 48000,
+                                      channels: AVAudioChannelCount(channels), interleaved: true)
+        else { throw CaptureError.deviceNotFound }
+        outFormat = out
+        converter = AVAudioConverter(from: inFormat, to: out)
+
+        input.installTap(onBus: 0, bufferSize: 960, format: inFormat) { [weak self] buf, _ in
+            self?.process(buf)
+        }
+        engine.prepare()
+        do { try engine.start() }
+        catch { throw CaptureError.engineStartFailed(-1) }
+    }
+
+    /// Stop capture and tear down the engine. Safe to call multiple times.
+    func stop() {
+        engine.inputNode.removeTap(onBus: 0)
+        if engine.isRunning { engine.stop() }
+        converter = nil
+        pending.removeAll(keepingCapacity: false)
+    }
+
+    // MARK: - Conversion (called on the tap's realtime thread)
+
+    private func process(_ inBuf: AVAudioPCMBuffer) {
+        guard let converter, let outFormat else { return }
+
+        // Output capacity must cover up-sampling (e.g. 44.1 → 48 kHz) plus slack.
+        let ratio = outFormat.sampleRate / inBuf.format.sampleRate
+        let cap = AVAudioFrameCount(Double(inBuf.frameLength) * ratio + 32)
+        guard cap > 0, let outBuf = AVAudioPCMBuffer(pcmFormat: outFormat, frameCapacity: cap)
+        else { return }
+
+        var fed = false
+        var err: NSError?
+        let status = converter.convert(to: outBuf, error: &err) { _, outStatus in
+            if fed { outStatus.pointee = .noDataNow; return nil }
+            fed = true
+            outStatus.pointee = .haveData
+            return inBuf
+        }
+        guard status != .error, outBuf.frameLength > 0,
+              let ch = outBuf.int16ChannelData else { return }
+
+        // Interleaved int16: all channels live in the first buffer (ch[0]).
+        let n = Int(outBuf.frameLength) * channels
+        pending.append(contentsOf: UnsafeBufferPointer(start: ch[0], count: n))
+        emit()
+    }
+
+    /// Fire `onPcm` for every whole 20 ms frame accumulated.
+    private func emit() {
+        let full = Self.frameSamplesPerChannel * channels
+        while pending.count >= full {
+            pending.withUnsafeBufferPointer { buf in
+                onPcm(buf.baseAddress!, Self.frameSamplesPerChannel, UInt32(channels))
+            }
+            pending.removeFirst(full)
+        }
+    }
+
+    // MARK: - UID → AudioDeviceID resolution
+
+    private static func deviceID(forUID uid: String) -> AudioDeviceID? {
+        var addr = AudioObjectPropertyAddress(
+            mSelector: kAudioHardwarePropertyTranslateUIDToDevice,
+            mScope: kAudioObjectPropertyScopeGlobal,
+            mElement: kAudioObjectPropertyElementMain)
+        var deviceID = AudioDeviceID(0)
+        var cfUID = uid as CFString
+        var size = UInt32(MemoryLayout<AudioDeviceID>.size)
+        let status = withUnsafeMutablePointer(to: &cfUID) { uidPtr -> OSStatus in
+            AudioObjectGetPropertyData(AudioObjectID(kAudioObjectSystemObject), &addr,
+                                       UInt32(MemoryLayout<CFString>.size), uidPtr,
+                                       &size, &deviceID)
+        }
+        return (status == noErr && deviceID != 0) ? deviceID : nil
+    }
+}
+
+// InputDeviceInfo / InputDeviceEnumerator — Core Audio input-device enumeration for the aux-stream
+// picker. Separate from the core's vc_list_devices (whose ids are miniaudio-opaque and can't be
+// passed to Core Audio); the aux device is client-captured, so the picker uses device UIDs.
+struct InputDeviceInfo: Equatable {
+    let uid: String       // stable across reboots/replugs — what we persist
+    let name: String
+    let isDefault: Bool
+}
+
+enum InputDeviceEnumerator {
+
+    /// All Core Audio devices that expose at least one input channel.
+    static func list() -> [InputDeviceInfo] {
+        let defaultUID = defaultInputUID()
+        var result: [InputDeviceInfo] = []
+
+        for devID in allDeviceIDs() {
+            guard inputChannelCount(devID) > 0 else { continue }
+            guard let uid = stringProperty(devID, kAudioDevicePropertyDeviceUID) else { continue }
+            let name = stringProperty(devID, kAudioObjectPropertyName)
+                ?? stringProperty(devID, kAudioDevicePropertyDeviceNameCFString)
+                ?? "Unknown input device"
+            result.append(InputDeviceInfo(uid: uid, name: name, isDefault: uid == defaultUID))
+        }
+        return result.sorted { $0.name.localizedCaseInsensitiveCompare($1.name) == .orderedAscending }
+    }
+
+    // MARK: - Core Audio helpers
+
+    private static func allDeviceIDs() -> [AudioDeviceID] {
+        var addr = AudioObjectPropertyAddress(
+            mSelector: kAudioHardwarePropertyDevices,
+            mScope: kAudioObjectPropertyScopeGlobal,
+            mElement: kAudioObjectPropertyElementMain)
+        var size = UInt32(0)
+        guard AudioObjectGetPropertyDataSize(AudioObjectID(kAudioObjectSystemObject), &addr, 0, nil,
+                                             &size) == noErr, size > 0 else { return [] }
+        let count = Int(size) / MemoryLayout<AudioDeviceID>.size
+        var ids = [AudioDeviceID](repeating: 0, count: count)
+        let status = AudioObjectGetPropertyData(AudioObjectID(kAudioObjectSystemObject), &addr, 0,
+                                                nil, &size, &ids)
+        return status == noErr ? ids : []
+    }
+
+    private static func inputChannelCount(_ devID: AudioDeviceID) -> Int {
+        var addr = AudioObjectPropertyAddress(
+            mSelector: kAudioDevicePropertyStreamConfiguration,
+            mScope: kAudioObjectPropertyScopeInput,
+            mElement: kAudioObjectPropertyElementMain)
+        var size = UInt32(0)
+        guard AudioObjectGetPropertyDataSize(devID, &addr, 0, nil, &size) == noErr, size > 0
+        else { return 0 }
+        let bufList = UnsafeMutableRawPointer.allocate(byteCount: Int(size),
+                                                       alignment: MemoryLayout<AudioBufferList>.alignment)
+        defer { bufList.deallocate() }
+        guard AudioObjectGetPropertyData(devID, &addr, 0, nil, &size, bufList) == noErr
+        else { return 0 }
+        let abl = UnsafeMutableAudioBufferListPointer(
+            bufList.assumingMemoryBound(to: AudioBufferList.self))
+        return abl.reduce(0) { $0 + Int($1.mNumberChannels) }
+    }
+
+    private static func defaultInputUID() -> String? {
+        var addr = AudioObjectPropertyAddress(
+            mSelector: kAudioHardwarePropertyDefaultInputDevice,
+            mScope: kAudioObjectPropertyScopeGlobal,
+            mElement: kAudioObjectPropertyElementMain)
+        var devID = AudioDeviceID(0)
+        var size = UInt32(MemoryLayout<AudioDeviceID>.size)
+        guard AudioObjectGetPropertyData(AudioObjectID(kAudioObjectSystemObject), &addr, 0, nil,
+                                         &size, &devID) == noErr, devID != 0 else { return nil }
+        return stringProperty(devID, kAudioDevicePropertyDeviceUID)
+    }
+
+    private static func stringProperty(_ devID: AudioDeviceID,
+                                       _ selector: AudioObjectPropertySelector) -> String? {
+        var addr = AudioObjectPropertyAddress(
+            mSelector: selector,
+            mScope: kAudioObjectPropertyScopeGlobal,
+            mElement: kAudioObjectPropertyElementMain)
+        var cf: CFString? = nil
+        var size = UInt32(MemoryLayout<CFString?>.size)
+        let status = withUnsafeMutablePointer(to: &cf) { ptr in
+            AudioObjectGetPropertyData(devID, &addr, 0, nil, &size, ptr)
+        }
+        guard status == noErr else { return nil }
+        return cf as String?
+    }
+}

clients/apple/macOS/VoiceCatMac/Windows/MainWindowController.swift

@@ -31,6 +31,8 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
     internal var micStreamId: UInt32 = 0
     private var screenStreamId: UInt32 = 0
     private var screenCapture: ScreenAudioCapture?
+    private var auxStreamId: UInt32 = 0          // 0 = aux (second input device) stream not active
+    private var auxCapture: InputDeviceCapture?  // client-side capture feeding the aux stream
     // Last app/exclusion choice from the share picker; reused as the default next time.
     private var screenAudioSelection: ScreenAudioSelection = .default
     internal var pttKeyCode: UInt16 = 0x60 {  // F8
@@ -60,6 +62,9 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
         static let vadThreshold = "voice.vadThreshold"
         static let inputGain    = "voice.inputGain"
         static let pttKeyCode   = "voice.pttKeyCode"
+        static let auxEnabled   = "voice.auxEnabled"
+        static let auxDeviceUID = "voice.auxDeviceUID"
+        static let auxGain      = "voice.auxGain"
     }
 
     internal var selectedInputMode: VoiceCatInputMode = .voiceActivation {
@@ -73,6 +78,20 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
     }
     internal var selectedInputDeviceId: String?
 
+    // Aux outgoing stream: a second hardware input device the client captures itself and feeds to
+    // the core (kind = AUX_DEVICE, external_feed). Device + volume only — aux is always-on (the
+    // core never gates AUX_DEVICE on VAD/PTT). auxDeviceUID is a Core Audio device UID (stable),
+    // NOT a core/miniaudio id. auxGain is read live by the capture feed, so the slider is instant.
+    internal var auxEnabled: Bool = false {
+        didSet { UserDefaults.standard.set(auxEnabled, forKey: AudioDefaults.auxEnabled) }
+    }
+    internal var auxDeviceUID: String? {
+        didSet { UserDefaults.standard.set(auxDeviceUID, forKey: AudioDefaults.auxDeviceUID) }
+    }
+    internal var auxGain: Float = 1.0 {
+        didSet { UserDefaults.standard.set(auxGain, forKey: AudioDefaults.auxGain) }
+    }
+
     // MARK: - UI components
     private let channelOutlineView = NSOutlineView()
     private let userTableView = NSTableView()
@@ -140,6 +159,13 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
         if d.object(forKey: AudioDefaults.pttKeyCode) != nil {
             pttKeyCode = UInt16(d.integer(forKey: AudioDefaults.pttKeyCode))
         }
+        auxEnabled = d.bool(forKey: AudioDefaults.auxEnabled)
+        if d.object(forKey: AudioDefaults.auxDeviceUID) != nil {
+            auxDeviceUID = d.string(forKey: AudioDefaults.auxDeviceUID)
+        }
+        if d.object(forKey: AudioDefaults.auxGain) != nil {
+            auxGain = d.float(forKey: AudioDefaults.auxGain)
+        }
     }
 
     required init?(coder: NSCoder) { fatalError() }
@@ -675,7 +701,8 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
         displayedUsers = []; userTableView.reloadData()
         users.removeAll(); talkingUsers.removeAll()
         stopScreenCapture()
-        currentChannelId = 0; micStreamId = 0; screenStreamId = 0
+        auxCapture?.stop(); auxCapture = nil  // connection gone — drop capture, no stopStream
+        currentChannelId = 0; micStreamId = 0; screenStreamId = 0; auxStreamId = 0
         composeField.isEnabled = false; sendButton.isEnabled = false
         joinVoiceButton?.isEnabled = false
         shareScreenButton?.isEnabled = false
@@ -757,10 +784,12 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
                 EventFeedback.shared.play(.voiceOn)
                 NSAccessibility.post(element: logTextView, notification: .announcementRequested,
                                      userInfo: [.announcement: "Joined voice", .priority: NSAccessibilityPriorityLevel.medium])
+                startAuxStream() // no-op unless the aux stream is enabled in settings
             } else {
                 addActivity("Failed to start microphone: \(result)")
             }
         } else {
+            stopAuxStream()
             client.setPushToTalk(false)
             client.stopStream(micStreamId)
             micStreamId = 0
@@ -771,6 +800,84 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
         }
     }
 
+    // MARK: - Aux input stream (second hardware input device)
+
+    /// Called by SettingsWindowController when the user toggles the aux checkbox.
+    func applyAuxEnabled(_ on: Bool) {
+        auxEnabled = on
+        guard micStreamId != 0 else { return } // not in voice — applied on next Join Voice
+        if on { startAuxStream() } else { stopAuxStream() }
+    }
+
+    /// Called by SettingsWindowController when the user picks a different aux device.
+    func applyAuxDevice(_ uid: String?) {
+        auxDeviceUID = uid
+        if auxStreamId != 0 { restartAuxCapture() }
+    }
+
+    private func startAuxStream() {
+        guard auxStreamId == 0, auxEnabled else { return }
+        let (result, streamId) = client.startStream(
+            StreamDescriptor(kind: .auxDevice, deviceId: nil, label: "Aux device", externalFeed: true))
+        guard result == .ok else {
+            addActivity("Failed to start aux stream: \(result)")
+            return
+        }
+        auxStreamId = streamId
+        startAuxCapture()
+        addActivity("Aux input stream active")
+    }
+
+    private func startAuxCapture() {
+        let capture = InputDeviceCapture(deviceUID: auxDeviceUID) { [weak self] ptr, spc, ch in
+            self?.feedAux(ptr, samplesPerChannel: spc, channels: ch)
+        }
+        auxCapture = capture
+        do { try capture.start() }
+        catch {
+            addActivity("Failed to open aux input device")
+            stopAuxStream()
+        }
+    }
+
+    // Re-open the capture on a different device while the aux stream stays up (the core stream id
+    // is unchanged — only the client-side capture source changes).
+    private func restartAuxCapture() {
+        guard auxStreamId != 0 else { return }
+        auxCapture?.stop()
+        auxCapture = nil
+        startAuxCapture()
+    }
+
+    private func stopAuxStream() {
+        auxCapture?.stop()
+        auxCapture = nil
+        if auxStreamId != 0 {
+            client.stopStream(auxStreamId)
+            auxStreamId = 0
+        }
+    }
+
+    // Fired on the capture's realtime thread. feedPcm is thread-safe. Gain is read live from
+    // auxGain each frame so the volume slider takes effect immediately.
+    private func feedAux(_ pcm: UnsafePointer<Int16>, samplesPerChannel: Int, channels: UInt32) {
+        guard auxStreamId != 0 else { return }
+        let gain = auxGain
+        if gain != 1.0 {
+            let n = samplesPerChannel * Int(channels)
+            var scaled = [Int16](repeating: 0, count: n)
+            for i in 0..<n {
+                let v = (Float(pcm[i]) * gain).rounded()
+                scaled[i] = Int16(max(-32768, min(32767, v)))
+            }
+            client.feedPcm(streamId: auxStreamId, pcm: scaled,
+                           samplesPerChannel: samplesPerChannel, channels: channels)
+        } else {
+            client.feedPcm(streamId: auxStreamId, pcm: pcm,
+                           samplesPerChannel: samplesPerChannel, channels: channels)
+        }
+    }
+
     /// Update toolbar button labels/states to reflect whether voice is active. Mirrors the
     /// Windows client's `SetVoiceJoinedState`.
     private func setVoiceJoinedState(_ joined: Bool) {
@@ -1158,6 +1265,7 @@ final class MainWindowController: NSWindowController, NSWindowDelegate {
         }
         client.setPushToTalk(false)
         stopScreenCapture()
+        if auxStreamId != 0 { stopAuxStream() }
         if screenStreamId != 0 { client.stopStream(screenStreamId) }
         if micStreamId != 0 { client.stopStream(micStreamId) }
         client.onLevel = nil

clients/apple/macOS/VoiceCatMac/Windows/SettingsWindowController.swift

@@ -58,6 +58,22 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
     }()
     private let inputGainValueLabel = NSTextField(labelWithString: "100%")
 
+    // Aux input stream: a second outgoing stream from another hardware input device (e.g. line-in
+    // / aux), captured client-side. Device + volume only — aux is always-on. Persisted via
+    // MainWindowController.auxEnabled / auxDeviceUID / auxGain.
+    private let auxCheckbox = NSButton(checkboxWithTitle: "Aux input stream (second device)",
+                                       target: nil, action: nil)
+    private let auxDevicePicker = NSPopUpButton()
+    private let auxRefreshButton = NSButton()
+    private let auxGainSlider: NSSlider = {
+        let s = NSSlider(value: 100, minValue: 0, maxValue: 300, target: nil, action: nil)
+        s.numberOfTickMarks = 0
+        return s
+    }()
+    private let auxGainValueLabel = NSTextField(labelWithString: "100%")
+    private let auxDeviceLabel = NSTextField(labelWithString: "Aux device:")
+    private let auxGainLabel = NSTextField(labelWithString: "Aux volume:")
+
     // Notification feedback controls. Read/write UserDefaults with the same keys VoiceCatCore's
     // FeedbackSettings reads, so EventFeedback honours these immediately.
     private let soundsCheckbox = NSButton(checkboxWithTitle: "Event sounds", target: nil, action: nil)
@@ -94,6 +110,7 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
         buildUI()
         syncFromMainController()
         loadInputDevices()
+        loadAuxDevices()
     }
 
     required init?(coder: NSCoder) { fatalError() }
@@ -175,6 +192,41 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
         inputGainRow.orientation = .horizontal
         inputGainRow.spacing = 8
 
+        // ── Aux input stream ──────────────────────────────────────────────────
+        let auxHeader = NSTextField(labelWithString: "Aux input stream")
+        auxHeader.font = .boldSystemFont(ofSize: NSFont.systemFontSize)
+
+        auxCheckbox.target = self
+        auxCheckbox.action = #selector(auxEnabledChanged)
+        auxCheckbox.setAccessibilityLabel("Enable aux input stream")
+        auxCheckbox.setAccessibilityHelp("Transmit a second hardware input device alongside your microphone.")
+
+        auxDeviceLabel.setAccessibilityLabel("Aux input device")
+        auxDevicePicker.setAccessibilityLabel("Aux input device")
+        auxDevicePicker.target = self
+        auxDevicePicker.action = #selector(auxDeviceChanged)
+        auxRefreshButton.title = "↺"
+        auxRefreshButton.bezelStyle = .rounded
+        auxRefreshButton.target = self
+        auxRefreshButton.action = #selector(refreshAuxDevicesClicked)
+        auxRefreshButton.setAccessibilityLabel("Refresh aux device list")
+        auxRefreshButton.toolTip = "Refresh"
+
+        auxGainLabel.setAccessibilityLabel("Aux volume")
+        auxGainSlider.target = self
+        auxGainSlider.action = #selector(auxGainChanged)
+        auxGainSlider.setAccessibilityLabel("Aux volume")
+        auxGainSlider.setAccessibilityHelp("Volume of the aux input stream. 100% is unity.")
+        auxGainValueLabel.setAccessibilityLabel("Aux volume value")
+
+        let auxDeviceRow = NSStackView(views: [auxDeviceLabel, auxDevicePicker, auxRefreshButton])
+        auxDeviceRow.orientation = .horizontal
+        auxDeviceRow.spacing = 8
+
+        let auxGainRow = NSStackView(views: [auxGainLabel, auxGainSlider, auxGainValueLabel])
+        auxGainRow.orientation = .horizontal
+        auxGainRow.spacing = 8
+
         // Notifications
         let notificationsHeader = NSTextField(labelWithString: "Notifications")
         notificationsHeader.font = .boldSystemFont(ofSize: NSFont.systemFontSize)
@@ -198,7 +250,8 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
         volumeRow.spacing = 8
 
         let stack = NSStackView(views: [inputModeRow, vadRow, inputGainRow, pttRow, deviceRow,
-                                        levelRow, notificationsHeader, soundsCheckbox, volumeRow,
+                                        levelRow, auxHeader, auxCheckbox, auxDeviceRow, auxGainRow,
+                                        notificationsHeader, soundsCheckbox, volumeRow,
                                         speechCheckbox, selfTalkCheckbox, pttSoundCheckbox])
         stack.orientation = .vertical
         stack.spacing = 12
@@ -218,6 +271,8 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
             levelMeter.widthAnchor.constraint(equalToConstant: 200),
             devicePicker.widthAnchor.constraint(greaterThanOrEqualToConstant: 180),
             soundsVolumeSlider.widthAnchor.constraint(greaterThanOrEqualToConstant: 200),
+            auxDevicePicker.widthAnchor.constraint(greaterThanOrEqualToConstant: 180),
+            auxGainSlider.widthAnchor.constraint(greaterThanOrEqualToConstant: 180),
         ])
 
         syncNotificationControls()
@@ -265,6 +320,11 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
         inputGainSlider.doubleValue = Double(mc.inputGain * 100)
         updateInputGainLabel()
 
+        auxCheckbox.state = mc.auxEnabled ? .on : .off
+        auxGainSlider.doubleValue = Double(mc.auxGain * 100)
+        updateAuxGainLabel()
+        updateAuxControlsEnabled()
+
         updateConditionalControls()
     }
 
@@ -338,6 +398,65 @@ final class SettingsWindowController: NSWindowController, NSWindowDelegate {
         }
     }
 
+    // MARK: - Aux input stream actions
+
+    @objc private func auxEnabledChanged() {
+        let on = auxCheckbox.state == .on
+        updateAuxControlsEnabled()
+        mainController?.applyAuxEnabled(on)
+    }
+
+    @objc private func refreshAuxDevicesClicked() { loadAuxDevices() }
+
+    @objc private func auxDeviceChanged() {
+        let uid = auxDevicePicker.selectedItem?.representedObject as? String
+        mainController?.applyAuxDevice(uid)
+    }
+
+    @objc private func auxGainChanged() {
+        mainController?.auxGain = Float(auxGainSlider.doubleValue) / 100.0
+        updateAuxGainLabel()
+    }
+
+    private func updateAuxGainLabel() {
+        let pct = Int(auxGainSlider.doubleValue.rounded())
+        auxGainValueLabel.stringValue = "\(pct)%"
+        auxGainSlider.setAccessibilityValue("\(pct) percent")
+    }
+
+    private func updateAuxControlsEnabled() {
+        let on = auxCheckbox.state == .on
+        auxDeviceLabel.isEnabled = on
+        auxDevicePicker.isEnabled = on
+        auxRefreshButton.isEnabled = on
+        auxGainLabel.isEnabled = on
+        auxGainSlider.isEnabled = on
+        auxGainValueLabel.isEnabled = on
+    }
+
+    private func loadAuxDevices() {
+        let devices = InputDeviceEnumerator.list()
+        let prevSelected = (auxDevicePicker.selectedItem?.representedObject as? String)
+            ?? mainController?.auxDeviceUID
+        auxDevicePicker.removeAllItems()
+        for d in devices {
+            let item = NSMenuItem(title: d.name, action: nil, keyEquivalent: "")
+            item.representedObject = d.uid
+            auxDevicePicker.menu?.addItem(item)
+        }
+        if let prev = prevSelected,
+           let item = auxDevicePicker.itemArray.first(where: { ($0.representedObject as? String) == prev }) {
+            auxDevicePicker.select(item)
+        } else if let def = devices.first(where: { $0.isDefault }),
+                  let item = auxDevicePicker.itemArray.first(where: { ($0.representedObject as? String) == def.uid }) {
+            auxDevicePicker.select(item)
+        } else if auxDevicePicker.numberOfItems > 0 {
+            auxDevicePicker.selectItem(at: 0)
+        }
+        // Persist the resolved selection so a relaunch (or Join Voice) opens the same device.
+        mainController?.auxDeviceUID = auxDevicePicker.selectedItem?.representedObject as? String
+    }
+
     // MARK: - Level meter (called by MainWindowController)
 
     func updateLevel(rms: Float) {

clients/windows/VoiceCat.App/Audio/InputDeviceCapture.cs

@@ -0,0 +1,454 @@
+using System.Runtime.InteropServices;
+
+// WASAPI shared-mode capture from a real hardware INPUT device (a microphone / line-in / aux
+// device), plus enumeration of capture endpoints for the aux-stream picker.
+//
+// This is the input-device analogue of ProcessLoopbackCapture (which captures *render* loopback
+// via the process-loopback activation hack). Here the source is an ordinary capture endpoint, so
+// we use the standard IMMDevice.Activate(IAudioClient) path with RCW interfaces — no vtable
+// gymnastics needed (a normal device's COM objects honour QueryInterface).
+//
+// Why client-side capture at all? The core already owns ONE capture device (the mic). It can't
+// open a second arbitrary input device, so for the aux stream the client captures the device and
+// feeds 48 kHz / 20 ms int16 frames into the core via vc_stream_feed_pcm — the same external-feed
+// pipeline screen-audio sharing uses. The device ids here are WASAPI endpoint ids and are NOT the
+// core's miniaudio ids, so the aux picker is populated independently of vc_list_devices.
+namespace VoiceCat.App.Audio;
+
+/// <summary>An audio input (capture) endpoint for the aux-stream device picker. <see cref="Id"/>
+/// is a WASAPI endpoint id (round-trip only; never construct by hand) — pass null to capture the
+/// system default. <see cref="ToString"/> returns the friendly name for ComboBox display.</summary>
+public sealed record InputDeviceInfo(string Id, string Name, bool IsDefault)
+{
+    public override string ToString() => Name;
+}
+
+/// <summary>Enumerates WASAPI capture endpoints. Separate from the core's vc_list_devices because
+/// the aux device is opened client-side and needs a WASAPI id, not a miniaudio one.</summary>
+public static class InputDeviceEnumerator
+{
+    public static IReadOnlyList<InputDeviceInfo> List()
+    {
+        var result = new List<InputDeviceInfo>();
+        InputDeviceCapture.IMMDeviceEnumerator? enumerator = null;
+        IntPtr collectionPtr = IntPtr.Zero;
+        string? defaultId = null;
+
+        try
+        {
+            enumerator = (InputDeviceCapture.IMMDeviceEnumerator)Activator.CreateInstance(
+                Type.GetTypeFromCLSID(new Guid("BCDE0395-E52F-467C-8E3D-C4579291692E"))!)!;
+
+            // Resolve the default capture endpoint id so the picker can flag it.
+            if (enumerator.GetDefaultAudioEndpoint(1 /*eCapture*/, 0 /*eConsole*/,
+                    out var defDev) == 0 && defDev != null)
+            {
+                try { if (defDev.GetId(out string id) == 0) defaultId = id; }
+                finally { Marshal.ReleaseComObject(defDev); }
+            }
+
+            // DEVICE_STATE_ACTIVE = 0x1 — only currently-usable endpoints.
+            if (enumerator.EnumAudioEndpoints(1 /*eCapture*/, 0x1, out collectionPtr) != 0
+                || collectionPtr == IntPtr.Zero)
+                return result;
+
+            var collection = (InputDeviceCapture.IMMDeviceCollection)
+                Marshal.GetObjectForIUnknown(collectionPtr);
+            collection.GetCount(out int count);
+
+            for (int i = 0; i < count; i++)
+            {
+                if (collection.Item(i, out var dev) != 0 || dev == null) continue;
+                try
+                {
+                    if (dev.GetId(out string id) != 0) continue;
+                    string name = ReadFriendlyName(dev) ?? "Unknown input device";
+                    result.Add(new InputDeviceInfo(id, name, id == defaultId));
+                }
+                finally { Marshal.ReleaseComObject(dev); }
+            }
+        }
+        catch { /* no audio subsystem / WASAPI unavailable — return what we have */ }
+        finally
+        {
+            if (collectionPtr != IntPtr.Zero) Marshal.Release(collectionPtr);
+            if (enumerator != null) Marshal.ReleaseComObject(enumerator);
+        }
+
+        return result.OrderBy(d => d.Name, StringComparer.OrdinalIgnoreCase).ToList();
+    }
+
+    private static string? ReadFriendlyName(InputDeviceCapture.IMMDevice dev)
+    {
+        if (dev.OpenPropertyStore(0 /*STGM_READ*/, out IntPtr storePtr) != 0
+            || storePtr == IntPtr.Zero)
+            return null;
+        try
+        {
+            var store = (InputDeviceCapture.IPropertyStore)Marshal.GetObjectForIUnknown(storePtr);
+            // PKEY_Device_FriendlyName = {a45c254e-df1c-4efd-8020-67d146a850e0}, pid 14.
+            var key = new InputDeviceCapture.PropertyKey
+            {
+                fmtid = new Guid("a45c254e-df1c-4efd-8020-67d146a850e0"),
+                pid = 14,
+            };
+            if (store.GetValue(ref key, out var pv) != 0) return null;
+            try
+            {
+                // VT_LPWSTR = 31.
+                return pv.vt == 31 ? Marshal.PtrToStringUni(pv.pointerValue) : null;
+            }
+            finally { InputDeviceCapture.PropVariantClear(ref pv); }
+        }
+        finally { Marshal.Release(storePtr); }
+    }
+}
+
+/// <summary>Captures a single hardware input device in WASAPI shared mode and raises a 20 ms
+/// (960 samples/channel @ 48 kHz, interleaved s16) frame event. The caller feeds these to the
+/// core's external-feed stream. Threading mirrors ProcessLoopbackCapture: all WASAPI work runs on
+/// a dedicated background (MTA) thread; the event fires on that thread.</summary>
+public sealed class InputDeviceCapture : IDisposable
+{
+    /// <summary>Fired on the capture thread every 20 ms: (interleaved s16 PCM, samplesPerChannel
+    /// = 960, channels).</summary>
+    public event Action<short[], int /*samplesPerChannel*/, int /*channels*/>? PcmFrameReady;
+
+    private const int SampleRate   = 48000;
+    private const int FrameSamples = 960;  // 20 ms
+
+    private readonly string? _deviceId;  // null = system default capture endpoint
+
+    private IAudioClient?        _audioClient;
+    private IAudioCaptureClient? _captureClient;
+
+    private AutoResetEvent? _bufferEvent;
+    private Thread?         _captureThread;
+    private volatile bool   _running;
+    private int             _channels;
+
+    // Accumulator: assembles driver-callback-sized fragments into FrameSamples chunks.
+    private short[] _accumBuf = [];
+    private int     _accumCount;
+
+    // Init-done signal: Set() by the capture thread after activation completes.
+    private readonly ManualResetEventSlim _initDone = new(false);
+    private bool _initOk;
+
+    public InputDeviceCapture(string? deviceId) => _deviceId = deviceId;
+
+    /// <summary>Starts capture. Blocks until WASAPI activation completes (typically &lt;100 ms).
+    /// Returns false if the device cannot be opened.</summary>
+    public bool Start()
+    {
+        if (_running) return false;
+        _running       = true;
+        _captureThread = new Thread(CaptureThreadProc)
+        {
+            IsBackground = true,
+            Name         = "AuxInputCapture",
+        };
+        _captureThread.Start();
+
+        bool ok = _initDone.Wait(5000) && _initOk;
+        if (!ok) _running = false;
+        return ok;
+    }
+
+    public void Stop()
+    {
+        _running = false;
+        _bufferEvent?.Set();
+        _captureThread?.Join(500);
+        try { _audioClient?.Stop(); } catch { /* device already gone */ }
+    }
+
+    public void Dispose()
+    {
+        Stop();
+        if (_captureClient != null) { Marshal.ReleaseComObject(_captureClient); _captureClient = null; }
+        if (_audioClient   != null) { Marshal.ReleaseComObject(_audioClient);   _audioClient   = null; }
+        _bufferEvent?.Dispose();
+        _initDone.Dispose();
+    }
+
+    // ── Capture thread (MTA) ──────────────────────────────────────────────────
+
+    private void CaptureThreadProc()
+    {
+        _initOk = ActivateAndStart();
+        _initDone.Set();
+        if (!_initOk) return;
+        CaptureLoop();
+    }
+
+    private bool ActivateAndStart()
+    {
+        if (!ActivateClient()) return false;
+
+        _bufferEvent = new AutoResetEvent(false);
+        if (_audioClient!.SetEventHandle(_bufferEvent.SafeWaitHandle.DangerousGetHandle()) < 0)
+            return false;
+
+        return _audioClient.Start() >= 0;
+    }
+
+    private bool ActivateClient()
+    {
+        IMMDeviceEnumerator? enumerator = null;
+        IMMDevice? device = null;
+        try
+        {
+            enumerator = (IMMDeviceEnumerator)Activator.CreateInstance(
+                Type.GetTypeFromCLSID(new Guid("BCDE0395-E52F-467C-8E3D-C4579291692E"))!)!;
+
+            int hr = _deviceId is null
+                ? enumerator.GetDefaultAudioEndpoint(1 /*eCapture*/, 0 /*eConsole*/, out device)
+                : enumerator.GetDevice(_deviceId, out device);
+            if (hr != 0 || device == null) return false;
+
+            var iidAudioClient = new Guid("1CB9AD4C-DBFA-4c32-B178-C2F568A703B2");
+            if (device.Activate(ref iidAudioClient, 0x17 /*CLSCTX_ALL*/, IntPtr.Zero,
+                    out object acObj) != 0 || acObj is not IAudioClient ac)
+                return false;
+            _audioClient = ac;
+
+            return InitializeStream();
+        }
+        catch { return false; }
+        finally
+        {
+            if (device     != null) Marshal.ReleaseComObject(device);
+            if (enumerator != null) Marshal.ReleaseComObject(enumerator);
+        }
+    }
+
+    private bool InitializeStream()
+    {
+        // Try s16 stereo first; fall back to s16 mono. AUTOCONVERTPCM lets the audio engine
+        // resample/convert the device's native format to our requested 48 kHz s16; EVENTCALLBACK
+        // drives the buffer-ready event. Shared mode (0), no LOOPBACK (this is a capture device).
+        //   AUDCLNT_STREAMFLAGS_EVENTCALLBACK    = 0x00040000
+        //   AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM   = 0x80000000
+        //   AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY = 0x08000000
+        const uint streamFlags = 0x00040000u | 0x80000000u | 0x08000000u;
+
+        foreach (int ch in new[] { 2, 1 })
+        {
+            var fmt = new WaveFormatEx
+            {
+                wFormatTag      = 1,           // WAVE_FORMAT_PCM
+                nChannels       = (ushort)ch,
+                nSamplesPerSec  = SampleRate,
+                wBitsPerSample  = 16,
+                nBlockAlign     = (ushort)(ch * 2),
+                nAvgBytesPerSec = (uint)(SampleRate * ch * 2),
+                cbSize          = 0,
+            };
+
+            IntPtr pFmt = Marshal.AllocHGlobal(Marshal.SizeOf<WaveFormatEx>());
+            try
+            {
+                Marshal.StructureToPtr(fmt, pFmt, false);
+                int hr = _audioClient!.Initialize(0 /*AUDCLNT_SHAREMODE_SHARED*/, streamFlags,
+                    2_000_000 /*200 ms hns*/, 0, pFmt, IntPtr.Zero);
+                if (hr >= 0)
+                {
+                    _channels   = ch;
+                    _accumBuf   = new short[FrameSamples * ch];
+                    _accumCount = 0;
+
+                    var iidCapture = new Guid("C8ADBD64-E71E-48a0-A4DE-185C395CD317");
+                    if (_audioClient.GetService(ref iidCapture, out object ccObj) != 0
+                        || ccObj is not IAudioCaptureClient cc)
+                        return false;
+                    _captureClient = cc;
+                    return true;
+                }
+                if (ch == 1) return false;
+            }
+            finally { Marshal.FreeHGlobal(pFmt); }
+        }
+        return false;
+    }
+
+    // ── Capture loop ─────────────────────────────────────────────────────────
+
+    private void CaptureLoop()
+    {
+        while (_running)
+        {
+            _bufferEvent!.WaitOne(100);
+            if (!_running) break;
+
+            while (_running)
+            {
+                if (_captureClient!.GetNextPacketSize(out uint packetSize) < 0 || packetSize == 0)
+                    break;
+
+                if (_captureClient.GetBuffer(out IntPtr dataPtr, out uint framesAvailable,
+                        out uint flags, out _, out _) < 0)
+                    break;
+
+                bool silent = (flags & 2) != 0; // AUDCLNT_BUFFERFLAGS_SILENT
+                if (framesAvailable > 0)
+                {
+                    if (silent) AccumulateSilence((int)framesAvailable);
+                    else        AccumulatePcm(dataPtr, (int)framesAvailable);
+                }
+
+                _captureClient.ReleaseBuffer(framesAvailable);
+            }
+        }
+    }
+
+    private unsafe void AccumulatePcm(IntPtr data, int frames)
+    {
+        var src   = (short*)data.ToPointer();
+        int total = frames * _channels;
+        int idx   = 0;
+        while (idx < total)
+        {
+            int space = _accumBuf.Length - _accumCount;
+            int copy  = Math.Min(total - idx, space);
+            fixed (short* dst = _accumBuf)
+                Buffer.MemoryCopy(src + idx, dst + _accumCount, copy * 2L, copy * 2L);
+            _accumCount += copy;
+            idx         += copy;
+            if (_accumCount == _accumBuf.Length)
+                FlushFrame();
+        }
+    }
+
+    private void AccumulateSilence(int frames)
+    {
+        int total = frames * _channels;
+        int idx   = 0;
+        while (idx < total)
+        {
+            int space = _accumBuf.Length - _accumCount;
+            int fill  = Math.Min(total - idx, space);
+            Array.Clear(_accumBuf, _accumCount, fill);
+            _accumCount += fill;
+            idx         += fill;
+            if (_accumCount == _accumBuf.Length)
+                FlushFrame();
+        }
+    }
+
+    private void FlushFrame()
+    {
+        var copy = new short[_accumBuf.Length];
+        _accumBuf.AsSpan().CopyTo(copy);
+        PcmFrameReady?.Invoke(copy, FrameSamples, _channels);
+        _accumCount = 0;
+    }
+
+    // ── COM declarations ───────────────────────────────────────────────────────
+    //
+    // Declared internal here so InputDeviceEnumerator can share them. These are standard MMDevice
+    // / WASAPI interfaces; a normal capture endpoint honours QueryInterface, so RCW marshalling is
+    // safe (unlike ProcessLoopbackCapture's process-loopback objects, which need raw vtable calls).
+
+    [ComImport, Guid("A95664D2-9614-4F35-A746-DE8DB63617E6"),
+     InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
+    internal interface IMMDeviceEnumerator
+    {
+        [PreserveSig] int EnumAudioEndpoints(int dataFlow, int stateMask, out IntPtr devices);
+        [PreserveSig] int GetDefaultAudioEndpoint(int dataFlow, int role, out IMMDevice endpoint);
+        [PreserveSig] int GetDevice([MarshalAs(UnmanagedType.LPWStr)] string id, out IMMDevice device);
+        [PreserveSig] int RegisterEndpointNotificationCallback(IntPtr client);
+        [PreserveSig] int UnregisterEndpointNotificationCallback(IntPtr client);
+    }
+
+    [ComImport, Guid("0BD7A1BE-7A1A-44DB-8397-CC5392387B5E"),
+     InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
+    internal interface IMMDeviceCollection
+    {
+        [PreserveSig] int GetCount(out int count);
+        [PreserveSig] int Item(int index, out IMMDevice device);
+    }
+
+    [ComImport, Guid("D666063F-1587-4E43-81F1-B948E807363F"),
+     InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
+    internal interface IMMDevice
+    {
+        [PreserveSig] int Activate(ref Guid iid, int clsCtx, IntPtr activationParams,
+            [MarshalAs(UnmanagedType.IUnknown)] out object ppInterface);
+        [PreserveSig] int OpenPropertyStore(int stgmAccess, out IntPtr propStore);
+        [PreserveSig] int GetId([MarshalAs(UnmanagedType.LPWStr)] out string id);
+        [PreserveSig] int GetState(out int state);
+    }
+
+    [ComImport, Guid("886D8EEB-8CF2-4446-8D02-CDBA1DBDCF99"),
+     InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
+    internal interface IPropertyStore
+    {
+        [PreserveSig] int GetCount(out int count);
+        [PreserveSig] int GetAt(int index, out PropertyKey key);
+        [PreserveSig] int GetValue(ref PropertyKey key, out PropVariant value);
+        [PreserveSig] int SetValue(ref PropertyKey key, ref PropVariant value);
+        [PreserveSig] int Commit();
+    }
+
+    [ComImport, Guid("1CB9AD4C-DBFA-4C32-B178-C2F568A703B2"),
+     InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
+    internal interface IAudioClient
+    {
+        [PreserveSig] int Initialize(int shareMode, uint streamFlags, long hnsBufferDuration,
+            long hnsPeriodicity, IntPtr pFormat, IntPtr audioSessionGuid);
+        [PreserveSig] int GetBufferSize(out uint numBufferFrames);
+        [PreserveSig] int GetStreamLatency(out long latency);
+        [PreserveSig] int GetCurrentPadding(out uint numPaddingFrames);
+        [PreserveSig] int IsFormatSupported(int shareMode, IntPtr pFormat, out IntPtr closestMatch);
+        [PreserveSig] int GetMixFormat(out IntPtr deviceFormat);
+        [PreserveSig] int GetDevicePeriod(out long defaultDevicePeriod, out long minimumDevicePeriod);
+        [PreserveSig] int Start();
+        [PreserveSig] int Stop();
+        [PreserveSig] int Reset();
+        [PreserveSig] int SetEventHandle(IntPtr eventHandle);
+        [PreserveSig] int GetService(ref Guid riid,
+            [MarshalAs(UnmanagedType.IUnknown)] out object ppv);
+    }
+
+    [ComImport, Guid("C8ADBD64-E71E-48A0-A4DE-185C395CD317"),
+     InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
+    internal interface IAudioCaptureClient
+    {
+        [PreserveSig] int GetBuffer(out IntPtr data, out uint numFramesToRead, out uint flags,
+            out ulong devicePosition, out ulong qpcPosition);
+        [PreserveSig] int ReleaseBuffer(uint numFramesRead);
+        [PreserveSig] int GetNextPacketSize(out uint numFramesInNextPacket);
+    }
+
+    [StructLayout(LayoutKind.Sequential, Pack = 2)]
+    private struct WaveFormatEx
+    {
+        public ushort wFormatTag;
+        public ushort nChannels;
+        public uint   nSamplesPerSec;
+        public uint   nAvgBytesPerSec;
+        public ushort nBlockAlign;
+        public ushort wBitsPerSample;
+        public ushort cbSize;
+    }
+
+    [StructLayout(LayoutKind.Sequential)]
+    internal struct PropertyKey
+    {
+        public Guid fmtid;
+        public int  pid;
+    }
+
+    // Minimal PROPVARIANT: we only ever read VT_LPWSTR (friendly name). x64 layout — the value
+    // union starts at offset 8 after vt(2)+reserved(6).
+    [StructLayout(LayoutKind.Explicit)]
+    internal struct PropVariant
+    {
+        [FieldOffset(0)] public ushort vt;
+        [FieldOffset(8)] public IntPtr pointerValue;
+    }
+
+    [DllImport("ole32.dll")]
+    internal static extern int PropVariantClear(ref PropVariant pvar);
+}

clients/windows/VoiceCat.App/Forms/AudioSettingsForm.cs

@@ -1,4 +1,5 @@
 using System.ComponentModel;
+using VoiceCat.App.Audio;
 using VoiceCat.App.Models;
 using VoiceCat.Interop;
 
@@ -15,12 +16,21 @@ public sealed class AudioSettingsForm : Form
     private readonly VoiceSettings _settings;
     private readonly uint _micStreamId;
 
+    // Aux-stream live-apply callbacks, supplied by MainForm (which owns the aux capture). Null
+    // when not connected — the controls still edit settings, just without live effect.
+    private readonly Action<bool>?    _applyAuxEnabled;
+    private readonly Action<string?>? _applyAuxDevice;
+    private readonly Action<float>?   _applyAuxGain;
+
     // Snapshot of original values so Cancel can restore them
     private readonly string? _origDeviceId;
     private readonly VcInputMode _origMode;
     private readonly int _origVadSlider;
     private readonly int _origMicGain;
     private readonly Keys _origPttKey;
+    private readonly bool _origAuxEnabled;
+    private readonly string? _origAuxDeviceId;
+    private readonly int _origAuxGain;
 
     private readonly ComboBox _cboDevice;
     private readonly Button _btnRefresh;
@@ -32,14 +42,25 @@ public sealed class AudioSettingsForm : Form
     private readonly Label _lblSensitivity;
     private readonly TrackBar _trkVad;
     private readonly TrackBar _trkGain;
+    private readonly CheckBox _chkAux;
+    private readonly Label _lblAuxDevice;
+    private readonly ComboBox _cboAuxDevice;
+    private readonly Button _btnAuxRefresh;
+    private readonly Label _lblAuxGain;
+    private readonly TrackBar _trkAuxGain;
 
     private Keys _pttKey;
 
-    public AudioSettingsForm(VoiceCatClient client, VoiceSettings settings, uint micStreamId)
+    public AudioSettingsForm(VoiceCatClient client, VoiceSettings settings, uint micStreamId,
+        Action<bool>? applyAuxEnabled = null, Action<string?>? applyAuxDevice = null,
+        Action<float>? applyAuxGain = null)
     {
         _client = client;
         _settings = settings;
         _micStreamId = micStreamId;
+        _applyAuxEnabled = applyAuxEnabled;
+        _applyAuxDevice = applyAuxDevice;
+        _applyAuxGain = applyAuxGain;
         _pttKey = (Keys)settings.PttKey;
 
         _origDeviceId = settings.InputDeviceId;
@@ -47,6 +68,9 @@ public sealed class AudioSettingsForm : Form
         _origVadSlider = settings.VadThresholdSlider;
         _origMicGain = settings.MicGain;
         _origPttKey = _pttKey;
+        _origAuxEnabled = settings.AuxEnabled;
+        _origAuxDeviceId = settings.AuxDeviceId;
+        _origAuxGain = settings.AuxGain;
 
         Text = "Audio settings";
         FormBorderStyle = FormBorderStyle.FixedDialog;
@@ -54,7 +78,7 @@ public sealed class AudioSettingsForm : Form
         MinimizeBox = false;
         StartPosition = FormStartPosition.CenterParent;
         AutoScaleMode = AutoScaleMode.Font;
-        ClientSize = new Size(420, 370);
+        ClientSize = new Size(420, 555);
 
         // ── Device row ────────────────────────────────────────────────────────
         var lblDevice = new Label
@@ -175,19 +199,83 @@ public sealed class AudioSettingsForm : Form
             "Boost a quiet microphone. 100 is unity gain; range 0–300 percent.";
         _trkGain.Scroll += TrkGain_Scroll;
 
+        // ── Aux input stream ──────────────────────────────────────────────────
+        // A second outgoing stream from another hardware input device (e.g. line-in / aux),
+        // captured client-side and fed to the core. Device + volume only — aux is always-on
+        // (the core never gates AUX_DEVICE on VAD/PTT).
+        _chkAux = new CheckBox
+        {
+            Text = "&Aux stream (second input device)",
+            Location = new Point(12, 348),
+            AutoSize = true,
+            Checked = settings.AuxEnabled,
+            AccessibleName = "Enable aux input stream",
+            AccessibleDescription =
+                "Transmit a second hardware input device alongside your microphone.",
+        };
+        _chkAux.CheckedChanged += ChkAux_CheckedChanged;
+
+        _lblAuxDevice = new Label
+        {
+            Text = "Aux d&evice:",
+            Location = new Point(12, 378),
+            AutoSize = true,
+        };
+        _cboAuxDevice = new ComboBox
+        {
+            Location = new Point(12, 398),
+            Width = 300,
+            DropDownStyle = ComboBoxStyle.DropDownList,
+            DisplayMember = "Name",
+            ValueMember = "Id",
+            AccessibleName = "Aux input device",
+            AccessibleDescription = "Select the second audio input device to transmit.",
+        };
+        _cboAuxDevice.SelectedIndexChanged += CboAuxDevice_SelectedIndexChanged;
+
+        _btnAuxRefresh = new Button
+        {
+            Text = "Re&fresh",
+            Location = new Point(320, 396),
+            Size = new Size(80, 26),
+        };
+        _btnAuxRefresh.Click += (_, _) => LoadAuxDevices();
+
+        _lblAuxGain = new Label
+        {
+            Text = "Aux vo&lume:",
+            Location = new Point(12, 434),
+            AutoSize = true,
+        };
+        _trkAuxGain = new TrackBar
+        {
+            Location = new Point(12, 454),
+            Size = new Size(200, 45),
+            Minimum = 0,
+            Maximum = 300,
+            TickFrequency = 25,
+            SmallChange = 5,
+            LargeChange = 25,
+            Value = Math.Clamp(settings.AuxGain, 0, 300),
+        };
+        _trkAuxGain.AccessibleName = "Aux volume";
+        _trkAuxGain.AccessibleDescription =
+            "Volume of the aux input stream. 100 is unity gain; range 0–300 percent.";
+        _trkAuxGain.Scroll += TrkAuxGain_Scroll;
+
         // ── OK / Cancel ───────────────────────────────────────────────────────
         var btnOk = new Button
         {
             Text = "&OK",
             DialogResult = DialogResult.OK,
-            Location = new Point(228, 334),
+            Location = new Point(228, 516),
             Size = new Size(80, 27),
         };
         var btnCancel = new Button
         {
             Text = "&Cancel",
             DialogResult = DialogResult.Cancel,
-            Location = new Point(316, 334),
+            Location = new Point(316, 516),
             Size = new Size(80, 27),
         };
 
@@ -205,6 +293,7 @@ public sealed class AudioSettingsForm : Form
             lblMode, _radioVad, _lblSensitivity, _trkVad,
             _radioPtt, _lblPttKey, _btnChangePtt, _radioAlwaysOn,
             lblGain, _trkGain,
+            _chkAux, _lblAuxDevice, _cboAuxDevice, _btnAuxRefresh, _lblAuxGain, _trkAuxGain,
             btnOk, btnCancel,
         ]);
 
@@ -217,6 +306,8 @@ public sealed class AudioSettingsForm : Form
         }
 
         LoadDevices();
+        LoadAuxDevices();
+        UpdateAuxControlsEnabled();
     }
 
     private void LoadDevices()
@@ -245,6 +336,65 @@ public sealed class AudioSettingsForm : Form
             _client.SetInputDevice(_micStreamId, dev.IsDefault ? null : dev.Id);
     }
 
+    // ── Aux input stream ──────────────────────────────────────────────────────
+
+    private void LoadAuxDevices()
+    {
+        string? currentId = (_cboAuxDevice.SelectedItem as InputDeviceInfo)?.Id
+                            ?? _settings.AuxDeviceId;
+
+        var devices = InputDeviceEnumerator.List().ToList();
+
+        // Set the data source AND restore the selection while unsubscribed so neither the initial
+        // populate nor a Refresh fires a spurious device-change (which would restart the capture).
+        _cboAuxDevice.SelectedIndexChanged -= CboAuxDevice_SelectedIndexChanged;
+        _cboAuxDevice.DataSource = new BindingList<InputDeviceInfo>(devices);
+
+        int idx = -1;
+        if (currentId is not null)
+            idx = devices.FindIndex(d => d.Id == currentId);
+        if (idx < 0)
+            idx = devices.FindIndex(d => d.IsDefault);
+        _cboAuxDevice.SelectedIndex = idx >= 0 ? idx : (devices.Count > 0 ? 0 : -1);
+        _cboAuxDevice.SelectedIndexChanged += CboAuxDevice_SelectedIndexChanged;
+
+        // Persist the resolved selection (null = default) so Join Voice opens the same device.
+        if (_cboAuxDevice.SelectedItem is InputDeviceInfo dev)
+            _settings.AuxDeviceId = dev.IsDefault ? null : dev.Id;
+    }
+
+    private void ChkAux_CheckedChanged(object? sender, EventArgs e)
+    {
+        _settings.AuxEnabled = _chkAux.Checked;
+        UpdateAuxControlsEnabled();
+        _applyAuxEnabled?.Invoke(_chkAux.Checked);
+    }
+
+    private void CboAuxDevice_SelectedIndexChanged(object? sender, EventArgs e)
+    {
+        if (_cboAuxDevice.SelectedItem is not InputDeviceInfo dev) return;
+        _settings.AuxDeviceId = dev.IsDefault ? null : dev.Id;
+        if (_chkAux.Checked)
+            _applyAuxDevice?.Invoke(dev.IsDefault ? null : dev.Id);
+    }
+
+    private void TrkAuxGain_Scroll(object? sender, EventArgs e)
+    {
+        _settings.AuxGain = _trkAuxGain.Value;
+        if (_chkAux.Checked)
+            _applyAuxGain?.Invoke(_trkAuxGain.Value / 100f);
+    }
+
+    private void UpdateAuxControlsEnabled()
+    {
+        bool on = _chkAux.Checked;
+        _lblAuxDevice.Enabled  = on;
+        _cboAuxDevice.Enabled  = on;
+        _btnAuxRefresh.Enabled = on;
+        _lblAuxGain.Enabled    = on;
+        _trkAuxGain.Enabled    = on;
+    }
+
     private void RadioMode_CheckedChanged(object? sender, EventArgs e)
     {
         var mode = CurrentMode();
@@ -310,6 +460,15 @@ public sealed class AudioSettingsForm : Form
                 _client.SetVadThreshold(0.1f * (1f - (_origVadSlider - 1f) / 99f));
             _client.SetInputGain(_origMicGain / 100f);
         }
+
+        // Aux: restore originals to settings and re-apply live (order: device + gain first so a
+        // re-enable starts with the right configuration).
+        _settings.AuxEnabled  = _origAuxEnabled;
+        _settings.AuxDeviceId = _origAuxDeviceId;
+        _settings.AuxGain     = _origAuxGain;
+        _applyAuxDevice?.Invoke(_origAuxDeviceId);
+        _applyAuxGain?.Invoke(_origAuxGain / 100f);
+        _applyAuxEnabled?.Invoke(_origAuxEnabled);
     }
 
     private void UpdatePttKeyLabel() =>

clients/windows/VoiceCat.App/Forms/MainForm.cs

@@ -28,6 +28,8 @@ public partial class MainForm : Form
     private uint _micStreamId; // 0 = not started
     private uint _screenStreamId; // 0 = not sharing screen audio
     private ProcessAudioMixer? _screenMixer; // non-null only in per-app capture mode
+    private uint _auxStreamId; // 0 = aux (second input device) stream not active
+    private InputDeviceCapture? _auxCapture; // client-side capture feeding the aux stream
     private Keys _pttKey = Keys.F8;
     private bool _pttEngaged; // guards the PTT cue against key-repeat
     private bool _serverMuted;
@@ -169,7 +171,16 @@ public partial class MainForm : Form
         var miAudio = new ToolStripMenuItem("&Audio...");
         miAudio.Click += (_, _) =>
         {
-            using var dlg = new AudioSettingsForm(_client, _voiceSettings, _micStreamId);
+            using var dlg = new AudioSettingsForm(_client, _voiceSettings, _micStreamId,
+                applyAuxEnabled: on =>
+                {
+                    if (_micStreamId == 0) return; // not in voice — applied on next Join Voice
+                    if (on) StartAuxStream(); else StopAuxStream();
+                },
+                applyAuxDevice: _ =>
+                {
+                    if (_auxStreamId != 0) RestartAuxCapture(); // settings.AuxDeviceId already updated
+                });
             dlg.ShowDialog(this);
             _pttKey = (Keys)_voiceSettings.PttKey;
         };
@@ -511,6 +522,7 @@ public partial class MainForm : Form
         _micStreamId = 0;
         _screenStreamId = 0;
         _screenMixer?.Stop(); _screenMixer?.Dispose(); _screenMixer = null;
+        DisposeAuxCapture(); _auxStreamId = 0; // connection gone — drop capture, no StopStream
         txtCompose.Enabled = false;
         btnSend.Enabled = false;
         tsbJoinVoice.Enabled = false;
@@ -641,6 +653,7 @@ public partial class MainForm : Form
                 SetVoiceJoinedState(true);
                 AddActivity("Joined voice — microphone active");
                 _feedback.PlaySound(SoundEvent.VoiceOn);
+                StartAuxStream(); // no-op unless the aux stream is enabled in settings
             }
             else
             {
@@ -649,6 +662,7 @@ public partial class MainForm : Form
         }
         else
         {
+            StopAuxStream();
             _client.SetPushToTalk(false);
             _client.StopStream(_micStreamId);
             _micStreamId = 0;
@@ -743,6 +757,82 @@ public partial class MainForm : Form
         AddActivity("Stopped sharing screen audio");
     }
 
+    // ── Aux input stream (second hardware input device) ─────────────────────────
+    // A second outgoing stream (kind = AUX_DEVICE, external_feed). The core can't open a second
+    // capture device, so we capture the chosen device here and feed PCM in — the same external-
+    // feed pipeline as per-app screen audio. Tied to the voice session: started on Join Voice
+    // (when enabled) and stopped on Leave Voice. The aux is always-on (the core never gates
+    // AUX_DEVICE on VAD/PTT); volume is applied client-side before feeding.
+
+    private void StartAuxStream()
+    {
+        if (_auxStreamId != 0 || !_voiceSettings.AuxEnabled) return;
+
+        var (result, streamId) = _client.StartStreamExternalFeed(VcStreamKind.AuxDevice, "Aux device");
+        if (result != VcResult.Ok)
+        {
+            AddActivity($"Failed to start aux stream: {result}");
+            return;
+        }
+        _auxStreamId = streamId;
+
+        _auxCapture = new InputDeviceCapture(_voiceSettings.AuxDeviceId);
+        _auxCapture.PcmFrameReady += OnAuxPcmFrame;
+        if (!_auxCapture.Start())
+        {
+            AddActivity("Failed to open aux input device");
+            StopAuxStream();
+            return;
+        }
+        AddActivity("Aux input stream active");
+    }
+
+    private void StopAuxStream()
+    {
+        DisposeAuxCapture();
+        if (_auxStreamId != 0)
+        {
+            _client.StopStream(_auxStreamId);
+            _auxStreamId = 0;
+        }
+    }
+
+    // Re-open the capture on a different device while the aux stream stays up (the core stream id
+    // is unchanged — only the client-side capture source changes).
+    private void RestartAuxCapture()
+    {
+        if (_auxStreamId == 0) return;
+        DisposeAuxCapture();
+        _auxCapture = new InputDeviceCapture(_voiceSettings.AuxDeviceId);
+        _auxCapture.PcmFrameReady += OnAuxPcmFrame;
+        if (!_auxCapture.Start())
+            AddActivity("Failed to open aux input device");
+    }
+
+    private void DisposeAuxCapture()
+    {
+        if (_auxCapture == null) return;
+        _auxCapture.PcmFrameReady -= OnAuxPcmFrame;
+        _auxCapture.Stop();
+        _auxCapture.Dispose();
+        _auxCapture = null;
+    }
+
+    // Fired on the capture thread. vc_stream_feed_pcm is thread-safe, so feed directly. Gain is
+    // read live from settings each frame (so the volume slider takes effect immediately).
+    private void OnAuxPcmFrame(short[] pcm, int samplesPerChannel, int channels)
+    {
+        if (_auxStreamId == 0) return;
+        float gain = _voiceSettings.AuxGain / 100f;
+        if (gain != 1f)
+        {
+            for (int i = 0; i < pcm.Length; i++)
+                pcm[i] = (short)Math.Clamp((int)MathF.Round(pcm[i] * gain),
+                    short.MinValue, short.MaxValue);
+        }
+        _client.StreamFeedPcm(_auxStreamId, pcm, samplesPerChannel, (uint)channels);
+    }
+
     private void TrkOutputVolume_Scroll(object? sender, EventArgs e) =>
         _client.SetOutputVolume(trkOutputVolume.Value / 100f);
 
@@ -1057,6 +1147,7 @@ public partial class MainForm : Form
         foreach (var win in _pmWindows.Values.ToList()) win.Close();
         _pmWindows.Clear();
         if (_screenStreamId != 0) { _screenMixer?.Stop(); _screenMixer?.Dispose(); _screenMixer = null; _client.StopStream(_screenStreamId); }
+        if (_auxStreamId != 0) StopAuxStream();
         if (_micStreamId != 0) _client.StopStream(_micStreamId);
         _client.Disconnect();
         _client.Dispose();

clients/windows/VoiceCat.App/Models/VoiceSettings.cs

@@ -27,6 +27,21 @@ public sealed class VoiceSettings
     /// <summary>Saved device ID from the last session; null means use the system default.</summary>
     public string? InputDeviceId { get; set; } = null;
 
+    /// <summary>Whether the secondary "aux" outgoing stream is enabled. The aux stream is a
+    /// second hardware input device the client captures itself and feeds to the core via
+    /// vc_stream_feed_pcm (kind = AUX_DEVICE, external_feed). Lets a user transmit e.g. mic +
+    /// a line-in at once.</summary>
+    public bool AuxEnabled { get; set; } = false;
+
+    /// <summary>WASAPI endpoint id of the aux capture device; null = system default capture
+    /// device. NOTE: this is a WASAPI device id (from <see cref="Audio.InputDeviceEnumerator"/>),
+    /// NOT a core/miniaudio id — the aux device is opened client-side, so the two id spaces differ.</summary>
+    public string? AuxDeviceId { get; set; } = null;
+
+    /// <summary>Aux input volume slider position, 0–300 percent (100 = unity). Applied client-side
+    /// to the captured PCM before feeding (the core's input gain is mic-only and global).</summary>
+    public int AuxGain { get; set; } = 100;
+
     private static readonly JsonSerializerOptions JsonOptions = new() { WriteIndented = true };
 
     private static string AppDataDir => Path.Combine(