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>

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) {