fix: capture system audio as mono

Meetingnotes.xcodeproj/project.pbxproj

@@ -276,7 +276,7 @@
 				CODE_SIGN_IDENTITY = "Apple Development";
 				CODE_SIGN_STYLE = Automatic;
 				COMBINE_HIDPI_IMAGES = YES;
-				CURRENT_PROJECT_VERSION = 37;
+				CURRENT_PROJECT_VERSION = 39;
 				DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
 				DEVELOPMENT_TEAM = G9LVHZAJNX;
 				ENABLE_HARDENED_RUNTIME = YES;
@@ -290,7 +290,7 @@
 					"@executable_path/../Frameworks",
 				);
 				MACOSX_DEPLOYMENT_TARGET = 15.0;
-				MARKETING_VERSION = 1.1.25;
+				MARKETING_VERSION = 1.1.27;
 				ONLY_ACTIVE_ARCH = NO;
 				OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
 				PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;
@@ -312,7 +312,7 @@
 				CODE_SIGN_IDENTITY = "Apple Development";
 				CODE_SIGN_STYLE = Automatic;
 				COMBINE_HIDPI_IMAGES = YES;
-				CURRENT_PROJECT_VERSION = 37;
+				CURRENT_PROJECT_VERSION = 39;
 				DEVELOPMENT_ASSET_PATHS = "\"meetingnotes/Preview Content\"";
 				DEVELOPMENT_TEAM = G9LVHZAJNX;
 				ENABLE_HARDENED_RUNTIME = YES;
@@ -326,7 +326,7 @@
 					"@executable_path/../Frameworks",
 				);
 				MACOSX_DEPLOYMENT_TARGET = 15.0;
-				MARKETING_VERSION = 1.1.25;
+				MARKETING_VERSION = 1.1.27;
 				ONLY_ACTIVE_ARCH = YES;
 				OTHER_SWIFT_FLAGS = "$(inherited) -D ENABLE_TCC_SPI";
 				PRODUCT_BUNDLE_IDENTIFIER = net.jamesbone.meetingnotes;

meetingnotes/ProcessTap/CoreAudioUtils.swift

@@ -116,6 +116,47 @@ extension AudioObjectID {
         try read(kAudioTapPropertyFormat, defaultValue: AudioStreamBasicDescription())
     }
 
+    /// Reads the virtual format of the first input stream exposed by this device.
+    ///
+    /// An aggregate device can adapt a tap to the active output hardware. Its
+    /// input stream format is therefore the format delivered to the I/O proc,
+    /// which can differ from the tap object's originally advertised format.
+    func readInputStreamBasicDescription() throws -> AudioStreamBasicDescription {
+        var streamsAddress = AudioObjectPropertyAddress(
+            mSelector: kAudioDevicePropertyStreams,
+            mScope: kAudioObjectPropertyScopeGlobal,
+            mElement: kAudioObjectPropertyElementMain
+        )
+        var dataSize: UInt32 = 0
+        var status = AudioObjectGetPropertyDataSize(self, &streamsAddress, 0, nil, &dataSize)
+        guard status == noErr else {
+            throw "Error reading device stream list size: \(status)"
+        }
+
+        var streamIDs = [AudioObjectID](
+            repeating: .unknown,
+            count: Int(dataSize) / MemoryLayout<AudioObjectID>.size
+        )
+        status = AudioObjectGetPropertyData(self, &streamsAddress, 0, nil, &dataSize, &streamIDs)
+        guard status == noErr else {
+            throw "Error reading device stream list: \(status)"
+        }
+
+        for streamID in streamIDs {
+            let direction: UInt32 = try streamID.read(
+                kAudioStreamPropertyDirection,
+                defaultValue: 0
+            )
+            guard direction == 1 else { continue }
+            return try streamID.read(
+                kAudioStreamPropertyVirtualFormat,
+                defaultValue: AudioStreamBasicDescription()
+            )
+        }
+
+        throw "Device has no input stream."
+    }
+
     private func requireSystemObject() throws {
         if self != .system { throw "Only supported for the system object." }
     }
@@ -307,4 +348,4 @@ extension AudioObjectID {
     func getDeviceName() throws -> String {
         return try readString(kAudioDevicePropertyDeviceNameCFString)
     }
-} 
+}

meetingnotes/ProcessTap/ProcessTap.swift

@@ -138,11 +138,11 @@ final class ProcessTap {
             tapDescription = CATapDescription(stereoMixdownOfProcesses: [process.objectID])
             logger.debug("Configuring tap for single process objectID: \(process.objectID)")
         case .systemAudio:
-            // Keep the HAL tap's buffer layout consistent with the default
-            // output stream. AudioManager performs the stereo-to-mono mix when
-            // it converts the captured audio to the 16 kHz transcription file.
-            tapDescription = CATapDescription(stereoGlobalTapButExcludeProcesses: [])
-            logger.debug("Configuring a stereo global system audio tap.")
+            // The transcription file is mono, so ask Core Audio for a mono
+            // mixdown at the source. This avoids interpreting a stereo HAL
+            // buffer as half as many frames before the 16 kHz conversion.
+            tapDescription = CATapDescription(monoGlobalTapButExcludeProcesses: [])
+            logger.info("Configuring a mono global system audio tap.")
         }
         
         tapDescription.uuid = UUID()
@@ -248,8 +248,20 @@ final class ProcessTap {
 
         do {
             logger.debug("Attempting to read audio tap stream basic description for tapID #\(tapID)...")
-            self.tapStreamDescription = try tapID.readAudioTapStreamBasicDescription()
-            logger.debug("Successfully read tap stream description: \(String(describing: self.tapStreamDescription))")
+            let advertisedDescription = try tapID.readAudioTapStreamBasicDescription()
+
+            // The aggregate device may adapt the tap to the active hardware's
+            // sample rate. Its input stream is the format actually delivered
+            // to the I/O proc, so use that rather than the tap's pre-aggregate
+            // advertisement. Using the latter can halve the written duration
+            // when, for example, a 48 kHz tap is delivered at 24 kHz.
+            do {
+                self.tapStreamDescription = try aggregateDeviceID.readInputStreamBasicDescription()
+                logger.info("Using aggregate input stream description: \(String(describing: self.tapStreamDescription), privacy: .public); tap advertised: \(String(describing: advertisedDescription), privacy: .public)")
+            } catch {
+                self.tapStreamDescription = advertisedDescription
+                logger.warning("Could not read aggregate input stream format; using tap format: \(error, privacy: .public)")
+            }
         } catch {
             logger.error("Failed to read audio tap stream basic description for tapID #\(tapID): \(error)")
             throw error // Propagate error