Audio input device selection

CHANGES.md

@@ -1,5 +1,10 @@
 # hear version history
 
+## v0.7 - 08/11/2025
+
+* Added support for specifying audio input device via `-n` flag
+* Added support for listing available audio input devices via `-a` flag
+
 ## v0.6 - 29/05/2025
 
 * Added timestamped output mode
@@ -9,15 +14,15 @@
 
 ## v0.5 - 05/11/2023
 
-* Now supports setting a timeout for speech recognition via the -t flag
+* Now supports setting a timeout for speech recognition via the `-t` flag
 
 ## v0.4 - 21/04/2023
 
-* Now supports -p flag to enable adding punctuation to speech recognition results (macOS 13+ only)
+* Now supports `-p` flag to enable adding punctuation to speech recognition results (macOS 13+ only)
 
 ## v0.3 - 25/03/2023
 
-* Now supports -x flag to specify "exit word", i.e. a word that causes the program to quit when heard
+* Now supports `-x` flag to specify "exit word", i.e. a word that causes the program to quit when heard
 
 ## v0.2 - 27/10/2022
 

Info.plist

@@ -2,19 +2,21 @@
 <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
 <plist version="1.0">
 <dict>
-    <key>CFBundleIdentifier</key>
-    <string>$(PRODUCT_BUNDLE_IDENTIFIER)</string>
-    <key>CFBundleShortVersionString</key>
-    <string>$(MARKETING_VERSION)</string>
-    <key>CFBundleSupportedPlatforms</key>
-    <array>
-    <string>MacOSX</string>
-    </array>
-    <key>LSMinimumSystemVersion</key>
-    <string>$(MACOSX_DEPLOYMENT_TARGET)</string>
-    <key>NSMicrophoneUsageDescription</key>
-    <string>This command line tool converts spoken audio to text.</string>
-    <key>NSSpeechRecognitionUsageDescription</key>
-    <string>This command line tool converts spoken audio to text.</string>
+	<key>CFBundleIdentifier</key>
+	<string>$(PRODUCT_BUNDLE_IDENTIFIER)</string>
+	<key>CFBundleShortVersionString</key>
+	<string>$(MARKETING_VERSION)</string>
+	<key>CFBundleSupportedPlatforms</key>
+	<array>
+		<string>MacOSX</string>
+	</array>
+	<key>LSMinimumSystemVersion</key>
+	<string>$(MACOSX_DEPLOYMENT_TARGET)</string>
+	<key>NSMicrophoneUsageDescription</key>
+	<string>This command line tool converts spoken audio to text.</string>
+	<key>NSCameraUseContinuityCameraDeviceType</key>
+	<true/>
+	<key>NSSpeechRecognitionUsageDescription</key>
+	<string>This command line tool converts spoken audio to text.</string>
 </dict>
 </plist>

Makefile

@@ -3,7 +3,7 @@
 XCODE_PROJ := "hear.xcodeproj"
 PROGRAM_NAME := "hear"
 BUILD_DIR := "products"
-VERSION := "0.6"
+VERSION := "0.7"
 
 all: clean build_unsigned
 

README.md

@@ -31,7 +31,7 @@ See the [man page](https://sveinbjorn.org/files/manpages/hear.1.html) for furthe
 [available](https://github.com/sveinbjornt/hear) under a [BSD license](#bsd-license).
 **If you find this program useful, please [make a donation](https://sveinbjorn.org/donations).**
 
-* **[⬇ Download hear 0.6](https://sveinbjorn.org/files/software/hear.zip)**
+* **[⬇ Download hear 0.7](https://sveinbjorn.org/files/software/hear.zip)**
 (~50 KB, ARM/Intel 64-bit, macOS 13 or later, Developer ID signed and notarized by Apple)
 
 ## Installation

hear.1

@@ -1,4 +1,4 @@
-.Dd May 29, 2025
+.Dd Nov 8, 2025
 .Dt HEAR 1
 .Os Darwin
 .Sh NAME
@@ -48,6 +48,10 @@ Set exit word. This causes the program to exit when a speech recognition result
 ends with the specified word.
 .It Fl t -timeout Ar seconds
 Exit if no recognition results are received within the specified number of seconds.
+.It Fl a -audio-input-devices
+List available audio input devices and exit.
+.It Fl n --input-device-id
+Specify ID of audio input device.
 .It Fl h -help
 Print help and exit.
 .It Fl v -version

hear.1.html

@@ -60,6 +60,12 @@
               Exit if no recognition results are received within the specified
               number of seconds.
 
+     <b>-a</b> <b>--audio-input-devices</b>
+              List available audio input devices and exit.
+
+     <b>-n</b> <b>--input-device-id</b>
+              Specify ID of audio input device.
+
      <b>-h</b> <b>--help</b>
               Print help and exit.
 
@@ -76,6 +82,6 @@
 
      Sveinbjorn Thordarson &lt;sveinbjorn@sveinbjorn.org&gt;
 
-Darwin                           May 29, 2025                           Darwin
+Darwin                         November 8, 2025                         Darwin
 </font></pre>
 </body></html>

src/Common.h

@@ -33,7 +33,7 @@
 #import <Foundation/Foundation.h>
 
 #define PROGRAM_NAME            @"hear"
-#define PROGRAM_VERSION         @"0.6"
+#define PROGRAM_VERSION         @"0.7"
 #define PROGRAM_AUTHOR          @"Sveinbjorn Thordarson"
 #define PROGRAM_AUTHOR_EMAIL    @"sveinbjorn@sveinbjorn.org"
 

src/Hear.h

@@ -44,8 +44,13 @@
                  addTimestamps:(BOOL)addTimestamps
                   subtitleMode:(BOOL)subtitle
                       exitWord:(NSString *)exitWord
-                       timeout:(CGFloat)timeout;
+                       timeout:(CGFloat)timeout
+                 inputDeviceID:(NSString *)inputDeviceID;
 
 + (void)printSupportedLocales;
++ (NSArray *)availableAudioInputDevices;
++ (BOOL)hasAvailableAudioInputDevice;
++ (BOOL)isAvailableAudioInputDevice:(NSString *)deviceID;
++ (void)printAvailableAudioInputDevices;
 
 @end

src/Hear.m

@@ -32,6 +32,7 @@
 
 #import "Hear.h"
 #import "Common.h"
+#import <CoreAudio/CoreAudio.h>
 
 @interface Hear()
 
@@ -53,6 +54,7 @@ @interface Hear()
 @property (nonatomic) BOOL subtitleMode;
 @property (nonatomic, retain) NSString *exitWord;
 @property (nonatomic) CGFloat timeout;
+@property (nonatomic, retain) NSString *inputDeviceID;
 
 @end
 
@@ -66,7 +68,9 @@ - (instancetype)initWithLocale:(NSString *)loc
                  addTimestamps:(BOOL)timestamps
                   subtitleMode:(BOOL)subtitle
                       exitWord:(NSString *)exitWord
-                       timeout:(CGFloat)timeout {
+                       timeout:(CGFloat)timeout
+                 inputDeviceID:(NSString *)inputDeviceID
+{
     if ((self = [super init])) {
 
         if ([[Hear supportedLocales] containsObject:loc] == NO) {
@@ -83,6 +87,7 @@ - (instancetype)initWithLocale:(NSString *)loc
         self.subtitleMode = subtitle;
         self.exitWord = exitWord;
         self.timeout = timeout;
+        self.inputDeviceID = inputDeviceID;
     }
     return self;
 }
@@ -144,7 +149,7 @@ - (void)initRecognizer {
 
     // Make sure recognition is available
     if (self.recognizer.isAvailable == NO) {
-        [self die:@"Speech recognizer not available. Try enabling Siri in System Preferences/Settings."];
+        [self die:@"Speech recognizer not available. Try enabling Siri in System Settings."];
     }
 
     if (self.useOnDeviceRecognition && !self.recognizer.supportsOnDeviceRecognition) {
@@ -301,6 +306,54 @@ - (void)processFileSubtitle {
 - (void)startListening {
     [self initRecognizer];
 
+    // Set the input device, if specified
+    if (self.inputDeviceID) {
+        AudioObjectPropertyAddress addr = {
+            kAudioHardwarePropertyDefaultInputDevice,
+            kAudioObjectPropertyScopeGlobal,
+            kAudioObjectPropertyElementMain
+        };
+
+        AudioDeviceID deviceID = kAudioObjectUnknown;
+
+        NSArray *devices = [Hear availableAudioInputDevices];
+        for (NSDictionary *device in devices) {
+            if ([device[@"id"] isEqualToString:self.inputDeviceID]) {
+
+                CFStringRef deviceUID = (__bridge CFStringRef)device[@"id"];
+
+                AudioValueTranslation value;
+                value.mInputData = &deviceUID;
+                value.mInputDataSize = sizeof(CFStringRef);
+                value.mOutputData = &deviceID;
+                value.mOutputDataSize = sizeof(AudioDeviceID);
+
+                UInt32 size = sizeof(AudioValueTranslation);
+
+                AudioObjectPropertyAddress addr = {
+                    kAudioHardwarePropertyDeviceForUID,
+                    kAudioObjectPropertyScopeGlobal,
+                    kAudioObjectPropertyElementMain
+                };
+
+                OSStatus status = AudioObjectGetPropertyData(kAudioObjectSystemObject, &addr, 0, NULL, &size, &value);
+                if (status != noErr) {
+                    [self die:@"Unable to get device ID for UID '%@'", self.inputDeviceID];
+                }
+                break;
+            }
+        }
+
+        if (deviceID == kAudioObjectUnknown) {
+            [self die:@"Audio input device with ID '%@' not found", self.inputDeviceID];
+        }
+
+        OSStatus status = AudioObjectSetPropertyData(kAudioObjectSystemObject, &addr, 0, NULL, sizeof(AudioDeviceID), &deviceID);
+        if (status != noErr) {
+            [self die:@"Error setting audio input device: %d", status];
+        }
+    }
+
     // Create speech recognition request
     self.request = [[SFSpeechAudioBufferRecognitionRequest alloc] init];
     if (self.request == nil) {
@@ -403,7 +456,7 @@ - (void)timedOut:(id)sender {
     exit(EXIT_SUCCESS);
 }
 
-#pragma mark - Class methods
+#pragma mark - Locales
 
 + (NSArray<NSString *> *)supportedLocales {
     NSMutableArray *localeIdentifiers = [NSMutableArray new];
@@ -418,6 +471,125 @@ + (void)printSupportedLocales {
     NSPrint([[Hear supportedLocales] componentsJoinedByString:@"\n"]);
 }
 
+#pragma mark - Audio Input Devices
+
++ (NSArray *)availableAudioInputDevices {
+    AudioObjectPropertyAddress addr = {
+        kAudioHardwarePropertyDevices,
+        kAudioObjectPropertyScopeGlobal,
+        kAudioObjectPropertyElementMain
+    };
+
+    UInt32 size;
+    OSStatus status = AudioObjectGetPropertyDataSize(kAudioObjectSystemObject, &addr, 0, NULL, &size);
+    if (status != noErr) {
+        return @[];
+    }
+
+    int count = size / sizeof(AudioDeviceID);
+    AudioDeviceID *deviceIDs = (AudioDeviceID *)malloc(size);
+    if (deviceIDs == NULL) {
+        return @[];
+    }
+
+    status = AudioObjectGetPropertyData(kAudioObjectSystemObject, &addr, 0, NULL, &size, deviceIDs);
+    if (status != noErr) {
+        free(deviceIDs);
+        return @[];
+    }
+
+    NSMutableArray *devices = [NSMutableArray array];
+
+    for (int i = 0; i < count; i++) {
+        AudioDeviceID deviceID = deviceIDs[i];
+
+        addr.mScope = kAudioDevicePropertyScopeInput;
+        addr.mSelector = kAudioDevicePropertyStreamConfiguration;
+        status = AudioObjectGetPropertyDataSize(deviceID, &addr, 0, NULL, &size);
+        if (status != noErr) {
+            continue;
+        }
+
+        AudioBufferList *bufferList = (AudioBufferList *)malloc(size);
+        status = AudioObjectGetPropertyData(deviceID, &addr, 0, NULL, &size, bufferList);
+        if (status != noErr) {
+            free(bufferList);
+            continue;
+        }
+
+        UInt32 channelCount = 0;
+        for (int j = 0; j < bufferList->mNumberBuffers; j++) {
+            channelCount += bufferList->mBuffers[j].mNumberChannels;
+        }
+        free(bufferList);
+
+        if (channelCount == 0) {
+            continue;
+        }
+
+        CFStringRef deviceName;
+        size = sizeof(deviceName);
+        addr.mSelector = kAudioDevicePropertyDeviceNameCFString;
+        status = AudioObjectGetPropertyData(deviceID, &addr, 0, NULL, &size, &deviceName);
+        if (status != noErr) {
+            continue;
+        }
+
+        CFStringRef deviceUID;
+        size = sizeof(deviceUID);
+        addr.mSelector = kAudioDevicePropertyDeviceUID;
+        status = AudioObjectGetPropertyData(deviceID, &addr, 0, NULL, &size, &deviceUID);
+        if (status != noErr) {
+            CFRelease(deviceName);
+            continue;
+        }
+
+        [devices addObject:@{
+            @"name": (__bridge NSString *)deviceName,
+            @"id": (__bridge NSString *)deviceUID
+        }];
+
+        CFRelease(deviceName);
+        CFRelease(deviceUID);
+    }
+
+    free(deviceIDs);
+
+    return devices;
+}
+
++ (BOOL)hasAvailableAudioInputDevice {
+    return [[Hear availableAudioInputDevices] count] != 0;
+}
+
++ (BOOL)isAvailableAudioInputDevice:(NSString *)deviceID {
+    NSArray *devices = [Hear availableAudioInputDevices];
+    for (NSDictionary *device in devices) {
+        if ([device[@"id"] isEqualToString:deviceID]) {
+            return YES;
+        }
+    }
+    return NO;
+}
+
++ (void)printAvailableAudioInputDevices {
+    NSArray *devices = [Hear availableAudioInputDevices];
+
+    if ([devices count] == 0) {
+        NSPrint(@"No audio input devices available");
+        return;
+    }
+
+    NSPrint(@"Available Audio Input Devices:");
+    NSUInteger num = 0;
+    for (NSDictionary *device in devices) {
+        num += 1;
+        NSPrint(@"%lu. %@ (ID: %@)", num, device[@"name"], device[@"id"]);
+    }
+}
+
+#pragma mark - Util
+
 + (BOOL)isFileSupportedByAVFoundation:(NSString *)filePath {
     // Create NSURL from file path
     NSURL *fileURL = [NSURL fileURLWithPath:filePath];