Merge pull request #19106 from SaracenOne/audio_mic

[WIP] Experimental microphone support

drivers/coreaudio/audio_driver_coreaudio.cpp

@@ -35,8 +35,23 @@
 #include "os/os.h"
 
 #define kOutputBus 0
+#define kInputBus 1
 
 #ifdef OSX_ENABLED
+OSStatus AudioDriverCoreAudio::input_device_address_cb(AudioObjectID inObjectID,
+		UInt32 inNumberAddresses, const AudioObjectPropertyAddress *inAddresses,
+		void *inClientData) {
+	AudioDriverCoreAudio *driver = (AudioDriverCoreAudio *)inClientData;
+
+	// If our selected device is the Default call set_device to update the
+	// kAudioOutputUnitProperty_CurrentDevice property
+	if (driver->capture_device_name == "Default") {
+		driver->capture_set_device("Default");
+	}
+
+	return noErr;
+}
+
 OSStatus AudioDriverCoreAudio::output_device_address_cb(AudioObjectID inObjectID,
 		UInt32 inNumberAddresses, const AudioObjectPropertyAddress *inAddresses,
 		void *inClientData) {
@@ -79,6 +94,11 @@ Error AudioDriverCoreAudio::init() {
 
 	result = AudioObjectAddPropertyListener(kAudioObjectSystemObject, &prop, &output_device_address_cb, this);
 	ERR_FAIL_COND_V(result != noErr, FAILED);
+
+	prop.mSelector = kAudioHardwarePropertyDefaultInputDevice;
+
+	result = AudioObjectAddPropertyListener(kAudioObjectSystemObject, &prop, &input_device_address_cb, this);
+	ERR_FAIL_COND_V(result != noErr, FAILED);
 #endif
 
 	AudioStreamBasicDescription strdesc;
@@ -102,6 +122,26 @@ Error AudioDriverCoreAudio::init() {
 			break;
 	}
 
+	zeromem(&strdesc, sizeof(strdesc));
+	size = sizeof(strdesc);
+	result = AudioUnitGetProperty(audio_unit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Output, kInputBus, &strdesc, &size);
+	ERR_FAIL_COND_V(result != noErr, FAILED);
+
+	switch (strdesc.mChannelsPerFrame) {
+		case 1: // Mono
+			capture_channels = 1;
+			break;
+
+		case 2: // Stereo
+			capture_channels = 2;
+			break;
+
+		default:
+			// Unknown number of channels, default to stereo
+			capture_channels = 2;
+			break;
+	}
+
 	mix_rate = GLOBAL_DEF_RST("audio/mix_rate", DEFAULT_MIX_RATE);
 
 	zeromem(&strdesc, sizeof(strdesc));
@@ -117,6 +157,11 @@ Error AudioDriverCoreAudio::init() {
 	result = AudioUnitSetProperty(audio_unit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Input, kOutputBus, &strdesc, sizeof(strdesc));
 	ERR_FAIL_COND_V(result != noErr, FAILED);
 
+	strdesc.mChannelsPerFrame = capture_channels;
+
+	result = AudioUnitSetProperty(audio_unit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Output, kInputBus, &strdesc, sizeof(strdesc));
+	ERR_FAIL_COND_V(result != noErr, FAILED);
+
 	int latency = GLOBAL_DEF_RST("audio/output_latency", DEFAULT_OUTPUT_LATENCY);
 	// Sample rate is independent of channels (ref: https://stackoverflow.com/questions/11048825/audio-sample-frequency-rely-on-channels)
 	buffer_frames = closest_power_of_2(latency * mix_rate / 1000);
@@ -126,8 +171,12 @@ Error AudioDriverCoreAudio::init() {
 	ERR_FAIL_COND_V(result != noErr, FAILED);
 #endif
 
-	buffer_size = buffer_frames * channels;
+	unsigned int buffer_size = buffer_frames * channels;
 	samples_in.resize(buffer_size);
+	input_buf.resize(buffer_size);
+	input_buffer.resize(buffer_size * 8);
+	input_position = 0;
+	input_size = 0;
 
 	if (OS::get_singleton()->is_stdout_verbose()) {
 		print_line("CoreAudio: detected " + itos(channels) + " channels");
@@ -141,6 +190,12 @@ Error AudioDriverCoreAudio::init() {
 	result = AudioUnitSetProperty(audio_unit, kAudioUnitProperty_SetRenderCallback, kAudioUnitScope_Input, kOutputBus, &callback, sizeof(callback));
 	ERR_FAIL_COND_V(result != noErr, FAILED);
 
+	zeromem(&callback, sizeof(AURenderCallbackStruct));
+	callback.inputProc = &AudioDriverCoreAudio::input_callback;
+	callback.inputProcRefCon = this;
+	result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_SetInputCallback, kAudioUnitScope_Global, 0, &callback, sizeof(callback));
+	ERR_FAIL_COND_V(result != noErr, FAILED);
+
 	result = AudioUnitInitialize(audio_unit);
 	ERR_FAIL_COND_V(result != noErr, FAILED);
 
@@ -192,6 +247,45 @@ OSStatus AudioDriverCoreAudio::output_callback(void *inRefCon,
 	return 0;
 };
 
+OSStatus AudioDriverCoreAudio::input_callback(void *inRefCon,
+		AudioUnitRenderActionFlags *ioActionFlags,
+		const AudioTimeStamp *inTimeStamp,
+		UInt32 inBusNumber, UInt32 inNumberFrames,
+		AudioBufferList *ioData) {
+
+	AudioDriverCoreAudio *ad = (AudioDriverCoreAudio *)inRefCon;
+	if (!ad->active) {
+		return 0;
+	}
+
+	ad->lock();
+
+	AudioBufferList bufferList;
+	bufferList.mNumberBuffers = 1;
+	bufferList.mBuffers[0].mData = ad->input_buf.ptrw();
+	bufferList.mBuffers[0].mNumberChannels = ad->capture_channels;
+	bufferList.mBuffers[0].mDataByteSize = ad->input_buf.size() * sizeof(int16_t);
+
+	OSStatus result = AudioUnitRender(ad->audio_unit, ioActionFlags, inTimeStamp, inBusNumber, inNumberFrames, &bufferList);
+	if (result == noErr) {
+		for (int i = 0; i < inNumberFrames * ad->capture_channels; i++) {
+			int32_t sample = ad->input_buf[i] << 16;
+			ad->input_buffer_write(sample);
+
+			if (ad->capture_channels == 1) {
+				// In case input device is single channel convert it to Stereo
+				ad->input_buffer_write(sample);
+			}
+		}
+	} else {
+		ERR_PRINT(("AudioUnitRender failed, code: " + itos(result)).utf8().get_data());
+	}
+
+	ad->unlock();
+
+	return result;
+}
+
 void AudioDriverCoreAudio::start() {
 	if (!active) {
 		OSStatus result = AudioOutputUnitStart(audio_unit);
@@ -222,9 +316,94 @@ AudioDriver::SpeakerMode AudioDriverCoreAudio::get_speaker_mode() const {
 	return get_speaker_mode_by_total_channels(channels);
 };
 
+void AudioDriverCoreAudio::lock() {
+	if (mutex)
+		mutex->lock();
+};
+
+void AudioDriverCoreAudio::unlock() {
+	if (mutex)
+		mutex->unlock();
+};
+
+bool AudioDriverCoreAudio::try_lock() {
+	if (mutex)
+		return mutex->try_lock() == OK;
+	return true;
+}
+
+void AudioDriverCoreAudio::finish() {
+	OSStatus result;
+
+	lock();
+
+	AURenderCallbackStruct callback;
+	zeromem(&callback, sizeof(AURenderCallbackStruct));
+	result = AudioUnitSetProperty(audio_unit, kAudioUnitProperty_SetRenderCallback, kAudioUnitScope_Input, kOutputBus, &callback, sizeof(callback));
+	if (result != noErr) {
+		ERR_PRINT("AudioUnitSetProperty failed");
+	}
+
+	if (active) {
+		result = AudioOutputUnitStop(audio_unit);
+		if (result != noErr) {
+			ERR_PRINT("AudioOutputUnitStop failed");
+		}
+
+		active = false;
+	}
+
+	result = AudioUnitUninitialize(audio_unit);
+	if (result != noErr) {
+		ERR_PRINT("AudioUnitUninitialize failed");
+	}
+
 #ifdef OSX_ENABLED
+	AudioObjectPropertyAddress prop;
+	prop.mSelector = kAudioHardwarePropertyDefaultOutputDevice;
+	prop.mScope = kAudioObjectPropertyScopeGlobal;
+	prop.mElement = kAudioObjectPropertyElementMaster;
 
-Array AudioDriverCoreAudio::get_device_list() {
+	result = AudioObjectRemovePropertyListener(kAudioObjectSystemObject, &prop, &output_device_address_cb, this);
+	if (result != noErr) {
+		ERR_PRINT("AudioObjectRemovePropertyListener failed");
+	}
+#endif
+
+	result = AudioComponentInstanceDispose(audio_unit);
+	if (result != noErr) {
+		ERR_PRINT("AudioComponentInstanceDispose failed");
+	}
+
+	unlock();
+
+	if (mutex) {
+		memdelete(mutex);
+		mutex = NULL;
+	}
+};
+
+Error AudioDriverCoreAudio::capture_start() {
+
+	UInt32 flag = 1;
+	OSStatus result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Input, kInputBus, &flag, sizeof(flag));
+	ERR_FAIL_COND_V(result != noErr, FAILED);
+
+	return OK;
+}
+
+Error AudioDriverCoreAudio::capture_stop() {
+
+	UInt32 flag = 0;
+	OSStatus result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Input, kInputBus, &flag, sizeof(flag));
+	ERR_FAIL_COND_V(result != noErr, FAILED);
+
+	return OK;
+}
+
+#ifdef OSX_ENABLED
+
+Array AudioDriverCoreAudio::_get_device_list(bool capture) {
 
 	Array list;
 
@@ -243,20 +422,20 @@ Array AudioDriverCoreAudio::get_device_list() {
 
 	UInt32 deviceCount = size / sizeof(AudioDeviceID);
 	for (UInt32 i = 0; i < deviceCount; i++) {
-		prop.mScope = kAudioDevicePropertyScopeOutput;
+		prop.mScope = capture ? kAudioDevicePropertyScopeInput : kAudioDevicePropertyScopeOutput;
 		prop.mSelector = kAudioDevicePropertyStreamConfiguration;
 
 		AudioObjectGetPropertyDataSize(audioDevices[i], &prop, 0, NULL, &size);
 		AudioBufferList *bufferList = (AudioBufferList *)malloc(size);
 		AudioObjectGetPropertyData(audioDevices[i], &prop, 0, NULL, &size, bufferList);
 
-		UInt32 outputChannelCount = 0;
+		UInt32 channelCount = 0;
 		for (UInt32 j = 0; j < bufferList->mNumberBuffers; j++)
-			outputChannelCount += bufferList->mBuffers[j].mNumberChannels;
+			channelCount += bufferList->mBuffers[j].mNumberChannels;
 
 		free(bufferList);
 
-		if (outputChannelCount >= 1) {
+		if (channelCount >= 1) {
 			CFStringRef cfname;
 
 			size = sizeof(CFStringRef);
@@ -281,21 +460,11 @@ Array AudioDriverCoreAudio::get_device_list() {
 	return list;
 }
 
-String AudioDriverCoreAudio::get_device() {
-
-	return device_name;
-}
-
-void AudioDriverCoreAudio::set_device(String device) {
-
-	device_name = device;
-	if (!active) {
-		return;
-	}
+void AudioDriverCoreAudio::_set_device(const String &device, bool capture) {
 
 	AudioDeviceID deviceId;
 	bool found = false;
-	if (device_name != "Default") {
+	if (device != "Default") {
 		AudioObjectPropertyAddress prop;
 
 		prop.mSelector = kAudioHardwarePropertyDevices;
@@ -309,20 +478,20 @@ void AudioDriverCoreAudio::set_device(String device) {
 
 		UInt32 deviceCount = size / sizeof(AudioDeviceID);
 		for (UInt32 i = 0; i < deviceCount && !found; i++) {
-			prop.mScope = kAudioDevicePropertyScopeOutput;
+			prop.mScope = capture ? kAudioDevicePropertyScopeInput : kAudioDevicePropertyScopeOutput;
 			prop.mSelector = kAudioDevicePropertyStreamConfiguration;
 
 			AudioObjectGetPropertyDataSize(audioDevices[i], &prop, 0, NULL, &size);
 			AudioBufferList *bufferList = (AudioBufferList *)malloc(size);
 			AudioObjectGetPropertyData(audioDevices[i], &prop, 0, NULL, &size, bufferList);
 
-			UInt32 outputChannelCount = 0;
+			UInt32 channelCount = 0;
 			for (UInt32 j = 0; j < bufferList->mNumberBuffers; j++)
-				outputChannelCount += bufferList->mBuffers[j].mNumberChannels;
+				channelCount += bufferList->mBuffers[j].mNumberChannels;
 
 			free(bufferList);
 
-			if (outputChannelCount >= 1) {
+			if (channelCount >= 1) {
 				CFStringRef cfname;
 
 				size = sizeof(CFStringRef);
@@ -335,7 +504,7 @@ void AudioDriverCoreAudio::set_device(String device) {
 				char *buffer = (char *)malloc(maxSize);
 				if (CFStringGetCString(cfname, buffer, maxSize, kCFStringEncodingUTF8)) {
 					String name = String(buffer) + " (" + itos(audioDevices[i]) + ")";
-					if (name == device_name) {
+					if (name == device) {
 						deviceId = audioDevices[i];
 						found = true;
 					}
@@ -351,7 +520,8 @@ void AudioDriverCoreAudio::set_device(String device) {
 	if (!found) {
 		// If we haven't found the desired device get the system default one
 		UInt32 size = sizeof(AudioDeviceID);
-		AudioObjectPropertyAddress property = { kAudioHardwarePropertyDefaultOutputDevice, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMaster };
+		UInt32 elem = capture ? kAudioHardwarePropertyDefaultInputDevice : kAudioHardwarePropertyDefaultOutputDevice;
+		AudioObjectPropertyAddress property = { elem, kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyElementMaster };
 
 		OSStatus result = AudioObjectGetPropertyData(kAudioObjectSystemObject, &property, 0, NULL, &size, &deviceId);
 		ERR_FAIL_COND(result != noErr);
@@ -360,79 +530,52 @@ void AudioDriverCoreAudio::set_device(String device) {
 	}
 
 	if (found) {
-		OSStatus result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, 0, &deviceId, sizeof(AudioDeviceID));
+		OSStatus result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, capture ? kInputBus : kOutputBus, &deviceId, sizeof(AudioDeviceID));
 		ERR_FAIL_COND(result != noErr);
+
+		// Reset audio input to keep synchronisation.
+		input_position = 0;
+		input_size = 0;
 	}
 }
 
-#endif
-
-void AudioDriverCoreAudio::lock() {
-	if (mutex)
-		mutex->lock();
-};
-
-void AudioDriverCoreAudio::unlock() {
-	if (mutex)
-		mutex->unlock();
-};
+Array AudioDriverCoreAudio::get_device_list() {
 
-bool AudioDriverCoreAudio::try_lock() {
-	if (mutex)
-		return mutex->try_lock() == OK;
-	return true;
+	return _get_device_list();
 }
 
-void AudioDriverCoreAudio::finish() {
-	OSStatus result;
+String AudioDriverCoreAudio::get_device() {
 
-	lock();
+	return device_name;
+}
 
-	AURenderCallbackStruct callback;
-	zeromem(&callback, sizeof(AURenderCallbackStruct));
-	result = AudioUnitSetProperty(audio_unit, kAudioUnitProperty_SetRenderCallback, kAudioUnitScope_Input, kOutputBus, &callback, sizeof(callback));
-	if (result != noErr) {
-		ERR_PRINT("AudioUnitSetProperty failed");
-	}
+void AudioDriverCoreAudio::set_device(String device) {
 
+	device_name = device;
 	if (active) {
-		result = AudioOutputUnitStop(audio_unit);
-		if (result != noErr) {
-			ERR_PRINT("AudioOutputUnitStop failed");
-		}
-
-		active = false;
+		_set_device(device_name);
 	}
+}
 
-	result = AudioUnitUninitialize(audio_unit);
-	if (result != noErr) {
-		ERR_PRINT("AudioUnitUninitialize failed");
+void AudioDriverCoreAudio::capture_set_device(const String &p_name) {
+
+	capture_device_name = p_name;
+	if (active) {
+		_set_device(capture_device_name, true);
 	}
+}
 
-#ifdef OSX_ENABLED
-	AudioObjectPropertyAddress prop;
-	prop.mSelector = kAudioHardwarePropertyDefaultOutputDevice;
-	prop.mScope = kAudioObjectPropertyScopeGlobal;
-	prop.mElement = kAudioObjectPropertyElementMaster;
+Array AudioDriverCoreAudio::capture_get_device_list() {
 
-	result = AudioObjectRemovePropertyListener(kAudioObjectSystemObject, &prop, &output_device_address_cb, this);
-	if (result != noErr) {
-		ERR_PRINT("AudioObjectRemovePropertyListener failed");
-	}
-#endif
+	return _get_device_list(true);
+}
 
-	result = AudioComponentInstanceDispose(audio_unit);
-	if (result != noErr) {
-		ERR_PRINT("AudioComponentInstanceDispose failed");
-	}
+String AudioDriverCoreAudio::capture_get_device() {
 
-	unlock();
+	return capture_device_name;
+}
 
-	if (mutex) {
-		memdelete(mutex);
-		mutex = NULL;
-	}
-};
+#endif
 
 AudioDriverCoreAudio::AudioDriverCoreAudio() {
 	active = false;
@@ -440,14 +583,15 @@ AudioDriverCoreAudio::AudioDriverCoreAudio() {
 
 	mix_rate = 0;
 	channels = 2;
+	capture_channels = 2;
 
-	buffer_size = 0;
 	buffer_frames = 0;
 
 	samples_in.clear();
 
 	device_name = "Default";
-};
+	capture_device_name = "Default";
+}
 
 AudioDriverCoreAudio::~AudioDriverCoreAudio(){};
 

drivers/coreaudio/audio_driver_coreaudio.h

@@ -48,15 +48,24 @@ class AudioDriverCoreAudio : public AudioDriver {
 	Mutex *mutex;
 
 	String device_name;
+	String capture_device_name;
 
 	int mix_rate;
 	unsigned int channels;
+	unsigned int capture_channels;
 	unsigned int buffer_frames;
-	unsigned int buffer_size;
 
 	Vector<int32_t> samples_in;
+	Vector<int16_t> input_buf;
 
 #ifdef OSX_ENABLED
+	Array _get_device_list(bool capture = false);
+	void _set_device(const String &device, bool capture = false);
+
+	static OSStatus input_device_address_cb(AudioObjectID inObjectID,
+			UInt32 inNumberAddresses, const AudioObjectPropertyAddress *inAddresses,
+			void *inClientData);
+
 	static OSStatus output_device_address_cb(AudioObjectID inObjectID,
 			UInt32 inNumberAddresses, const AudioObjectPropertyAddress *inAddresses,
 			void *inClientData);
@@ -68,6 +77,12 @@ class AudioDriverCoreAudio : public AudioDriver {
 			UInt32 inBusNumber, UInt32 inNumberFrames,
 			AudioBufferList *ioData);
 
+	static OSStatus input_callback(void *inRefCon,
+			AudioUnitRenderActionFlags *ioActionFlags,
+			const AudioTimeStamp *inTimeStamp,
+			UInt32 inBusNumber, UInt32 inNumberFrames,
+			AudioBufferList *ioData);
+
 public:
 	const char *get_name() const {
 		return "CoreAudio";
@@ -77,18 +92,27 @@ public:
 	virtual void start();
 	virtual int get_mix_rate() const;
 	virtual SpeakerMode get_speaker_mode() const;
-#ifdef OSX_ENABLED
-	virtual Array get_device_list();
-	virtual String get_device();
-	virtual void set_device(String device);
-#endif
+
 	virtual void lock();
 	virtual void unlock();
 	virtual void finish();
 
+	virtual Error capture_start();
+	virtual Error capture_stop();
+
 	bool try_lock();
 	void stop();
 
+#ifdef OSX_ENABLED
+	virtual Array get_device_list();
+	virtual String get_device();
+	virtual void set_device(String device);
+
+	virtual Array capture_get_device_list();
+	virtual void capture_set_device(const String &p_name);
+	virtual String capture_get_device();
+#endif
+
 	AudioDriverCoreAudio();
 	~AudioDriverCoreAudio();
 };

drivers/pulseaudio/audio_driver_pulseaudio.cpp

@@ -64,18 +64,32 @@ void AudioDriverPulseAudio::pa_sink_info_cb(pa_context *c, const pa_sink_info *l
 	ad->pa_status++;
 }
 
+void AudioDriverPulseAudio::pa_source_info_cb(pa_context *c, const pa_source_info *l, int eol, void *userdata) {
+	AudioDriverPulseAudio *ad = (AudioDriverPulseAudio *)userdata;
+
+	// If eol is set to a positive number, you're at the end of the list
+	if (eol > 0) {
+		return;
+	}
+
+	ad->pa_rec_map = l->channel_map;
+	ad->pa_status++;
+}
+
 void AudioDriverPulseAudio::pa_server_info_cb(pa_context *c, const pa_server_info *i, void *userdata) {
 	AudioDriverPulseAudio *ad = (AudioDriverPulseAudio *)userdata;
 
+	ad->capture_default_device = i->default_source_name;
 	ad->default_device = i->default_sink_name;
 	ad->pa_status++;
 }
 
-void AudioDriverPulseAudio::detect_channels() {
+void AudioDriverPulseAudio::detect_channels(bool capture) {
 
-	pa_channel_map_init_stereo(&pa_map);
+	pa_channel_map_init_stereo(capture ? &pa_rec_map : &pa_map);
 
-	if (device_name == "Default") {
+	String device = capture ? capture_device_name : device_name;
+	if (device == "Default") {
 		// Get the default output device name
 		pa_status = 0;
 		pa_operation *pa_op = pa_context_get_server_info(pa_ctx, &AudioDriverPulseAudio::pa_server_info_cb, (void *)this);
@@ -93,16 +107,22 @@ void AudioDriverPulseAudio::detect_channels() {
 		}
 	}
 
-	char device[1024];
-	if (device_name == "Default") {
-		strcpy(device, default_device.utf8().get_data());
+	char dev[1024];
+	if (device == "Default") {
+		strcpy(dev, capture ? capture_default_device.utf8().get_data() : default_device.utf8().get_data());
 	} else {
-		strcpy(device, device_name.utf8().get_data());
+		strcpy(dev, device.utf8().get_data());
 	}
 
 	// Now using the device name get the amount of channels
 	pa_status = 0;
-	pa_operation *pa_op = pa_context_get_sink_info_by_name(pa_ctx, device, &AudioDriverPulseAudio::pa_sink_info_cb, (void *)this);
+	pa_operation *pa_op;
+	if (capture) {
+		pa_op = pa_context_get_source_info_by_name(pa_ctx, dev, &AudioDriverPulseAudio::pa_source_info_cb, (void *)this);
+	} else {
+		pa_op = pa_context_get_sink_info_by_name(pa_ctx, dev, &AudioDriverPulseAudio::pa_sink_info_cb, (void *)this);
+	}
+
 	if (pa_op) {
 		while (pa_status == 0) {
 			int ret = pa_mainloop_iterate(pa_ml, 1, NULL);
@@ -113,7 +133,11 @@ void AudioDriverPulseAudio::detect_channels() {
 
 		pa_operation_unref(pa_op);
 	} else {
-		ERR_PRINT("pa_context_get_sink_info_by_name error");
+		if (capture) {
+			ERR_PRINT("pa_context_get_source_info_by_name error");
+		} else {
+			ERR_PRINT("pa_context_get_sink_info_by_name error");
+		}
 	}
 }
 
@@ -195,6 +219,10 @@ Error AudioDriverPulseAudio::init_device() {
 	samples_in.resize(buffer_frames * channels);
 	samples_out.resize(pa_buffer_size);
 
+	// Reset audio input to keep synchronisation.
+	input_position = 0;
+	input_size = 0;
+
 	return OK;
 }
 
@@ -287,74 +315,71 @@ float AudioDriverPulseAudio::get_latency() {
 void AudioDriverPulseAudio::thread_func(void *p_udata) {
 
 	AudioDriverPulseAudio *ad = (AudioDriverPulseAudio *)p_udata;
+	unsigned int write_ofs = 0;
+	size_t avail_bytes = 0;
 
 	while (!ad->exit_thread) {
 
-		ad->lock();
-		ad->start_counting_ticks();
-
-		if (!ad->active) {
-			for (unsigned int i = 0; i < ad->pa_buffer_size; i++) {
-				ad->samples_out.write[i] = 0;
-			}
+		size_t read_bytes = 0;
+		size_t written_bytes = 0;
 
-		} else {
-			ad->audio_server_process(ad->buffer_frames, ad->samples_in.ptrw());
+		if (avail_bytes == 0) {
+			ad->lock();
+			ad->start_counting_ticks();
 
-			if (ad->channels == ad->pa_map.channels) {
+			if (!ad->active) {
 				for (unsigned int i = 0; i < ad->pa_buffer_size; i++) {
-					ad->samples_out.write[i] = ad->samples_in[i] >> 16;
+					ad->samples_out.write[i] = 0;
 				}
 			} else {
-				// Uneven amount of channels
-				unsigned int in_idx = 0;
-				unsigned int out_idx = 0;
+				ad->audio_server_process(ad->buffer_frames, ad->samples_in.ptrw());
 
-				for (unsigned int i = 0; i < ad->buffer_frames; i++) {
-					for (unsigned int j = 0; j < ad->pa_map.channels - 1; j++) {
-						ad->samples_out.write[out_idx++] = ad->samples_in[in_idx++] >> 16;
+				if (ad->channels == ad->pa_map.channels) {
+					for (unsigned int i = 0; i < ad->pa_buffer_size; i++) {
+						ad->samples_out.write[i] = ad->samples_in[i] >> 16;
+					}
+				} else {
+					// Uneven amount of channels
+					unsigned int in_idx = 0;
+					unsigned int out_idx = 0;
+
+					for (unsigned int i = 0; i < ad->buffer_frames; i++) {
+						for (unsigned int j = 0; j < ad->pa_map.channels - 1; j++) {
+							ad->samples_out.write[out_idx++] = ad->samples_in[in_idx++] >> 16;
+						}
+						uint32_t l = ad->samples_in[in_idx++];
+						uint32_t r = ad->samples_in[in_idx++];
+						ad->samples_out.write[out_idx++] = (l >> 1 + r >> 1) >> 16;
 					}
-					uint32_t l = ad->samples_in[in_idx++];
-					uint32_t r = ad->samples_in[in_idx++];
-					ad->samples_out.write[out_idx++] = (l >> 1 + r >> 1) >> 16;
 				}
 			}
+
+			avail_bytes = ad->pa_buffer_size * sizeof(int16_t);
+			write_ofs = 0;
+			ad->stop_counting_ticks();
+			ad->unlock();
 		}
 
-		int error_code;
-		int byte_size = ad->pa_buffer_size * sizeof(int16_t);
+		ad->lock();
+		ad->start_counting_ticks();
+
 		int ret;
 		do {
 			ret = pa_mainloop_iterate(ad->pa_ml, 0, NULL);
 		} while (ret > 0);
 
-		if (pa_stream_get_state(ad->pa_str) == PA_STREAM_READY) {
-			const void *ptr = ad->samples_out.ptr();
-			while (byte_size > 0) {
-				size_t bytes = pa_stream_writable_size(ad->pa_str);
-				if (bytes > 0) {
-					if (bytes > byte_size) {
-						bytes = byte_size;
-					}
-
-					ret = pa_stream_write(ad->pa_str, ptr, bytes, NULL, 0LL, PA_SEEK_RELATIVE);
-					if (ret >= 0) {
-						byte_size -= bytes;
-						ptr = (const char *)ptr + bytes;
-					}
+		if (avail_bytes > 0 && pa_stream_get_state(ad->pa_str) == PA_STREAM_READY) {
+			size_t bytes = pa_stream_writable_size(ad->pa_str);
+			if (bytes > 0) {
+				size_t bytes_to_write = MIN(bytes, avail_bytes);
+				const void *ptr = ad->samples_out.ptr();
+				ret = pa_stream_write(ad->pa_str, ptr + write_ofs, bytes_to_write, NULL, 0LL, PA_SEEK_RELATIVE);
+				if (ret != 0) {
+					ERR_PRINT("pa_stream_write error");
 				} else {
-					ret = pa_mainloop_iterate(ad->pa_ml, 0, NULL);
-					if (ret == 0) {
-						// If pa_mainloop_iterate returns 0 sleep for 1 msec to wait
-						// for the stream to be able to process more bytes
-						ad->stop_counting_ticks();
-						ad->unlock();
-
-						OS::get_singleton()->delay_usec(1000);
-
-						ad->lock();
-						ad->start_counting_ticks();
-					}
+					avail_bytes -= bytes_to_write;
+					write_ofs += bytes_to_write;
+					written_bytes += bytes_to_write;
 				}
 			}
 		}
@@ -379,8 +404,64 @@ void AudioDriverPulseAudio::thread_func(void *p_udata) {
 			}
 		}
 
+		if (ad->pa_rec_str && pa_stream_get_state(ad->pa_rec_str) == PA_STREAM_READY) {
+			size_t bytes = pa_stream_readable_size(ad->pa_rec_str);
+			if (bytes > 0) {
+				const void *ptr = NULL;
+				size_t maxbytes = ad->input_buffer.size() * sizeof(int16_t);
+
+				bytes = MIN(bytes, maxbytes);
+				ret = pa_stream_peek(ad->pa_rec_str, &ptr, &bytes);
+				if (ret != 0) {
+					ERR_PRINT("pa_stream_peek error");
+				} else {
+					int16_t *srcptr = (int16_t *)ptr;
+					for (size_t i = bytes >> 1; i > 0; i--) {
+						int32_t sample = int32_t(*srcptr++) << 16;
+						ad->input_buffer_write(sample);
+
+						if (ad->pa_rec_map.channels == 1) {
+							// In case input device is single channel convert it to Stereo
+							ad->input_buffer_write(sample);
+						}
+					}
+
+					read_bytes += bytes;
+					ret = pa_stream_drop(ad->pa_rec_str);
+					if (ret != 0) {
+						ERR_PRINT("pa_stream_drop error");
+					}
+				}
+			}
+
+			// User selected a new device, finish the current one so we'll init the new device
+			if (ad->capture_device_name != ad->capture_new_device) {
+				ad->capture_device_name = ad->capture_new_device;
+				ad->capture_finish_device();
+
+				Error err = ad->capture_init_device();
+				if (err != OK) {
+					ERR_PRINT("PulseAudio: capture_init_device error");
+					ad->capture_device_name = "Default";
+					ad->capture_new_device = "Default";
+
+					err = ad->capture_init_device();
+					if (err != OK) {
+						ad->active = false;
+						ad->exit_thread = true;
+						break;
+					}
+				}
+			}
+		}
+
 		ad->stop_counting_ticks();
 		ad->unlock();
+
+		// Let the thread rest a while if we haven't read or write anything
+		if (written_bytes == 0 && read_bytes == 0) {
+			OS::get_singleton()->delay_usec(1000);
+		}
 	}
 
 	ad->thread_exited = true;
@@ -510,11 +591,165 @@ void AudioDriverPulseAudio::finish() {
 	thread = NULL;
 }
 
+Error AudioDriverPulseAudio::capture_init_device() {
+
+	// If there is a specified device check that it is really present
+	if (capture_device_name != "Default") {
+		Array list = capture_get_device_list();
+		if (list.find(capture_device_name) == -1) {
+			capture_device_name = "Default";
+			capture_new_device = "Default";
+		}
+	}
+
+	detect_channels(true);
+	switch (pa_rec_map.channels) {
+		case 1: // Mono
+		case 2: // Stereo
+			break;
+
+		default:
+			WARN_PRINTS("PulseAudio: Unsupported number of input channels: " + itos(pa_rec_map.channels));
+			pa_channel_map_init_stereo(&pa_rec_map);
+			break;
+	}
+
+	if (OS::get_singleton()->is_stdout_verbose()) {
+		print_line("PulseAudio: detected " + itos(pa_rec_map.channels) + " input channels");
+	}
+
+	pa_sample_spec spec;
+
+	spec.format = PA_SAMPLE_S16LE;
+	spec.channels = pa_rec_map.channels;
+	spec.rate = mix_rate;
+
+	int latency = 30;
+	input_buffer_frames = closest_power_of_2(latency * mix_rate / 1000);
+	int buffer_size = input_buffer_frames * spec.channels;
+
+	pa_buffer_attr attr;
+	attr.fragsize = buffer_size * sizeof(int16_t);
+
+	pa_rec_str = pa_stream_new(pa_ctx, "Record", &spec, &pa_rec_map);
+	if (pa_rec_str == NULL) {
+		ERR_PRINTS("PulseAudio: pa_stream_new error: " + String(pa_strerror(pa_context_errno(pa_ctx))));
+		ERR_FAIL_V(ERR_CANT_OPEN);
+	}
+
+	const char *dev = capture_device_name == "Default" ? NULL : capture_device_name.utf8().get_data();
+	pa_stream_flags flags = pa_stream_flags(PA_STREAM_INTERPOLATE_TIMING | PA_STREAM_ADJUST_LATENCY | PA_STREAM_AUTO_TIMING_UPDATE);
+	int error_code = pa_stream_connect_record(pa_rec_str, dev, &attr, flags);
+	if (error_code < 0) {
+		ERR_PRINTS("PulseAudio: pa_stream_connect_record error: " + String(pa_strerror(error_code)));
+		ERR_FAIL_V(ERR_CANT_OPEN);
+	}
+
+	input_buffer.resize(input_buffer_frames * 8);
+	input_position = 0;
+	input_size = 0;
+
+	return OK;
+}
+
+void AudioDriverPulseAudio::capture_finish_device() {
+
+	if (pa_rec_str) {
+		int ret = pa_stream_disconnect(pa_rec_str);
+		if (ret != 0) {
+			ERR_PRINTS("PulseAudio: pa_stream_disconnect error: " + String(pa_strerror(ret)));
+		}
+		pa_stream_unref(pa_rec_str);
+		pa_rec_str = NULL;
+	}
+}
+
+Error AudioDriverPulseAudio::capture_start() {
+
+	lock();
+	Error err = capture_init_device();
+	unlock();
+
+	return err;
+}
+
+Error AudioDriverPulseAudio::capture_stop() {
+	lock();
+	capture_finish_device();
+	unlock();
+
+	return OK;
+}
+
+void AudioDriverPulseAudio::capture_set_device(const String &p_name) {
+
+	lock();
+	capture_new_device = p_name;
+	unlock();
+}
+
+void AudioDriverPulseAudio::pa_sourcelist_cb(pa_context *c, const pa_source_info *l, int eol, void *userdata) {
+	AudioDriverPulseAudio *ad = (AudioDriverPulseAudio *)userdata;
+
+	// If eol is set to a positive number, you're at the end of the list
+	if (eol > 0) {
+		return;
+	}
+
+	if (l->monitor_of_sink == PA_INVALID_INDEX) {
+		ad->pa_rec_devices.push_back(l->name);
+	}
+
+	ad->pa_status++;
+}
+
+Array AudioDriverPulseAudio::capture_get_device_list() {
+
+	pa_rec_devices.clear();
+	pa_rec_devices.push_back("Default");
+
+	if (pa_ctx == NULL) {
+		return pa_rec_devices;
+	}
+
+	lock();
+
+	// Get the device list
+	pa_status = 0;
+	pa_operation *pa_op = pa_context_get_source_info_list(pa_ctx, pa_sourcelist_cb, (void *)this);
+	if (pa_op) {
+		while (pa_status == 0) {
+			int ret = pa_mainloop_iterate(pa_ml, 1, NULL);
+			if (ret < 0) {
+				ERR_PRINT("pa_mainloop_iterate error");
+			}
+		}
+
+		pa_operation_unref(pa_op);
+	} else {
+		ERR_PRINT("pa_context_get_server_info error");
+	}
+
+	unlock();
+
+	return pa_rec_devices;
+}
+
+String AudioDriverPulseAudio::capture_get_device() {
+
+	lock();
+	String name = capture_device_name;
+	unlock();
+
+	return name;
+}
+
 AudioDriverPulseAudio::AudioDriverPulseAudio() {
 
 	pa_ml = NULL;
 	pa_ctx = NULL;
 	pa_str = NULL;
+	pa_rec_str = NULL;
 
 	mutex = NULL;
 	thread = NULL;
@@ -528,6 +763,7 @@ AudioDriverPulseAudio::AudioDriverPulseAudio() {
 
 	mix_rate = 0;
 	buffer_frames = 0;
+	input_buffer_frames = 0;
 	pa_buffer_size = 0;
 	channels = 0;
 	pa_ready = 0;

drivers/pulseaudio/audio_driver_pulseaudio.h

@@ -47,22 +47,30 @@ class AudioDriverPulseAudio : public AudioDriver {
 	pa_mainloop *pa_ml;
 	pa_context *pa_ctx;
 	pa_stream *pa_str;
+	pa_stream *pa_rec_str;
 	pa_channel_map pa_map;
+	pa_channel_map pa_rec_map;
 
 	String device_name;
 	String new_device;
 	String default_device;
 
+	String capture_device_name;
+	String capture_new_device;
+	String capture_default_device;
+
 	Vector<int32_t> samples_in;
 	Vector<int16_t> samples_out;
 
 	unsigned int mix_rate;
 	unsigned int buffer_frames;
+	unsigned int input_buffer_frames;
 	unsigned int pa_buffer_size;
 	int channels;
 	int pa_ready;
 	int pa_status;
 	Array pa_devices;
+	Array pa_rec_devices;
 
 	bool active;
 	bool thread_exited;
@@ -72,13 +80,18 @@ class AudioDriverPulseAudio : public AudioDriver {
 
 	static void pa_state_cb(pa_context *c, void *userdata);
 	static void pa_sink_info_cb(pa_context *c, const pa_sink_info *l, int eol, void *userdata);
+	static void pa_source_info_cb(pa_context *c, const pa_source_info *l, int eol, void *userdata);
 	static void pa_server_info_cb(pa_context *c, const pa_server_info *i, void *userdata);
 	static void pa_sinklist_cb(pa_context *c, const pa_sink_info *l, int eol, void *userdata);
+	static void pa_sourcelist_cb(pa_context *c, const pa_source_info *l, int eol, void *userdata);
 
 	Error init_device();
 	void finish_device();
 
-	void detect_channels();
+	Error capture_init_device();
+	void capture_finish_device();
+
+	void detect_channels(bool capture = false);
 
 	static void thread_func(void *p_udata);
 
@@ -91,15 +104,24 @@ public:
 	virtual void start();
 	virtual int get_mix_rate() const;
 	virtual SpeakerMode get_speaker_mode() const;
+
 	virtual Array get_device_list();
 	virtual String get_device();
 	virtual void set_device(String device);
+
+	virtual Array capture_get_device_list();
+	virtual void capture_set_device(const String &p_name);
+	virtual String capture_get_device();
+
 	virtual void lock();
 	virtual void unlock();
 	virtual void finish();
 
 	virtual float get_latency();
 
+	virtual Error capture_start();
+	virtual Error capture_stop();
+
 	AudioDriverPulseAudio();
 	~AudioDriverPulseAudio();
 };

drivers/wasapi/audio_driver_wasapi.cpp

@@ -32,6 +32,8 @@
 
 #include "audio_driver_wasapi.h"
 
+#include <Functiondiscoverykeys_devpkey.h>
+
 #include "os/os.h"
 #include "project_settings.h"
 
@@ -52,8 +54,22 @@ const CLSID CLSID_MMDeviceEnumerator = __uuidof(MMDeviceEnumerator);
 const IID IID_IMMDeviceEnumerator = __uuidof(IMMDeviceEnumerator);
 const IID IID_IAudioClient = __uuidof(IAudioClient);
 const IID IID_IAudioRenderClient = __uuidof(IAudioRenderClient);
+const IID IID_IAudioCaptureClient = __uuidof(IAudioCaptureClient);
+
+#define SAFE_RELEASE(memory) \
+	if ((memory) != NULL) {  \
+		(memory)->Release(); \
+		(memory) = NULL;     \
+	}
 
-static bool default_device_changed = false;
+#define REFTIMES_PER_SEC 10000000
+#define REFTIMES_PER_MILLISEC 10000
+
+#define CAPTURE_BUFFER_CHANNELS 2
+
+static StringName capture_device_id;
+static bool default_render_device_changed = false;
+static bool default_capture_device_changed = false;
 
 class CMMNotificationClient : public IMMNotificationClient {
 	LONG _cRef;
@@ -109,8 +125,13 @@ public:
 	}
 
 	HRESULT STDMETHODCALLTYPE OnDefaultDeviceChanged(EDataFlow flow, ERole role, LPCWSTR pwstrDeviceId) {
-		if (flow == eRender && role == eConsole) {
-			default_device_changed = true;
+		if (role == eConsole) {
+			if (flow == eRender) {
+				default_render_device_changed = true;
+			} else if (flow == eCapture) {
+				default_capture_device_changed = true;
+				capture_device_id = String(pwstrDeviceId);
+			}
 		}
 
 		return S_OK;
@@ -123,7 +144,7 @@ public:
 
 static CMMNotificationClient notif_client;
 
-Error AudioDriverWASAPI::init_device(bool reinit) {
+Error AudioDriverWASAPI::audio_device_init(AudioDeviceWASAPI *p_device, bool p_capture, bool reinit) {
 
 	WAVEFORMATEX *pwfex;
 	IMMDeviceEnumerator *enumerator = NULL;
@@ -134,12 +155,12 @@ Error AudioDriverWASAPI::init_device(bool reinit) {
 	HRESULT hr = CoCreateInstance(CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, IID_IMMDeviceEnumerator, (void **)&enumerator);
 	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
-	if (device_name == "Default") {
-		hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, &device);
+	if (p_device->device_name == "Default") {
+		hr = enumerator->GetDefaultAudioEndpoint(p_capture ? eCapture : eRender, eConsole, &device);
 	} else {
 		IMMDeviceCollection *devices = NULL;
 
-		hr = enumerator->EnumAudioEndpoints(eRender, DEVICE_STATE_ACTIVE, &devices);
+		hr = enumerator->EnumAudioEndpoints(p_capture ? eCapture : eRender, DEVICE_STATE_ACTIVE, &devices);
 		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
 		LPWSTR strId = NULL;
@@ -165,7 +186,7 @@ Error AudioDriverWASAPI::init_device(bool reinit) {
 			hr = props->GetValue(PKEY_Device_FriendlyName, &propvar);
 			ERR_BREAK(hr != S_OK);
 
-			if (device_name == String(propvar.pwszVal)) {
+			if (p_device->device_name == String(propvar.pwszVal)) {
 				hr = device->GetId(&strId);
 				ERR_BREAK(hr != S_OK);
 
@@ -186,9 +207,10 @@ Error AudioDriverWASAPI::init_device(bool reinit) {
 		}
 
 		if (device == NULL) {
-			hr = enumerator->GetDefaultAudioEndpoint(eRender, eConsole, &device);
+			hr = enumerator->GetDefaultAudioEndpoint(p_capture ? eCapture : eRender, eConsole, &device);
 		}
 	}
+
 	if (reinit) {
 		// In case we're trying to re-initialize the device prevent throwing this error on the console,
 		// otherwise if there is currently no device available this will spam the console.
@@ -200,11 +222,15 @@ Error AudioDriverWASAPI::init_device(bool reinit) {
 	}
 
 	hr = enumerator->RegisterEndpointNotificationCallback(&notif_client);
+	SAFE_RELEASE(enumerator)
+
 	if (hr != S_OK) {
 		ERR_PRINT("WASAPI: RegisterEndpointNotificationCallback error");
 	}
 
-	hr = device->Activate(IID_IAudioClient, CLSCTX_ALL, NULL, (void **)&audio_client);
+	hr = device->Activate(IID_IAudioClient, CLSCTX_ALL, NULL, (void **)&p_device->audio_client);
+	SAFE_RELEASE(device)
+
 	if (reinit) {
 		if (hr != S_OK) {
 			return ERR_CANT_OPEN;
@@ -213,75 +239,89 @@ Error AudioDriverWASAPI::init_device(bool reinit) {
 		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 	}
 
-	hr = audio_client->GetMixFormat(&pwfex);
+	hr = p_device->audio_client->GetMixFormat(&pwfex);
 	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
 	// Since we're using WASAPI Shared Mode we can't control any of these, we just tag along
-	wasapi_channels = pwfex->nChannels;
-	format_tag = pwfex->wFormatTag;
-	bits_per_sample = pwfex->wBitsPerSample;
+	p_device->channels = pwfex->nChannels;
+	p_device->format_tag = pwfex->wFormatTag;
+	p_device->bits_per_sample = pwfex->wBitsPerSample;
+	p_device->frame_size = (p_device->bits_per_sample / 8) * p_device->channels;
 
-	switch (wasapi_channels) {
-		case 2: // Stereo
-		case 4: // Surround 3.1
-		case 6: // Surround 5.1
-		case 8: // Surround 7.1
-			channels = wasapi_channels;
-			break;
-
-		default:
-			WARN_PRINTS("WASAPI: Unsupported number of channels: " + itos(wasapi_channels));
-			channels = 2;
-			break;
-	}
-
-	if (format_tag == WAVE_FORMAT_EXTENSIBLE) {
+	if (p_device->format_tag == WAVE_FORMAT_EXTENSIBLE) {
 		WAVEFORMATEXTENSIBLE *wfex = (WAVEFORMATEXTENSIBLE *)pwfex;
 
 		if (wfex->SubFormat == KSDATAFORMAT_SUBTYPE_PCM) {
-			format_tag = WAVE_FORMAT_PCM;
+			p_device->format_tag = WAVE_FORMAT_PCM;
 		} else if (wfex->SubFormat == KSDATAFORMAT_SUBTYPE_IEEE_FLOAT) {
-			format_tag = WAVE_FORMAT_IEEE_FLOAT;
+			p_device->format_tag = WAVE_FORMAT_IEEE_FLOAT;
 		} else {
 			ERR_PRINT("WASAPI: Format not supported");
 			ERR_FAIL_V(ERR_CANT_OPEN);
 		}
 	} else {
-		if (format_tag != WAVE_FORMAT_PCM && format_tag != WAVE_FORMAT_IEEE_FLOAT) {
+		if (p_device->format_tag != WAVE_FORMAT_PCM && p_device->format_tag != WAVE_FORMAT_IEEE_FLOAT) {
 			ERR_PRINT("WASAPI: Format not supported");
 			ERR_FAIL_V(ERR_CANT_OPEN);
 		}
 	}
 
-	DWORD streamflags = AUDCLNT_STREAMFLAGS_EVENTCALLBACK;
+	DWORD streamflags = 0;
 	if (mix_rate != pwfex->nSamplesPerSec) {
 		streamflags |= AUDCLNT_STREAMFLAGS_RATEADJUST;
 		pwfex->nSamplesPerSec = mix_rate;
 		pwfex->nAvgBytesPerSec = pwfex->nSamplesPerSec * pwfex->nChannels * (pwfex->wBitsPerSample / 8);
 	}
 
-	hr = audio_client->Initialize(AUDCLNT_SHAREMODE_SHARED, streamflags, 0, 0, pwfex, NULL);
+	hr = p_device->audio_client->Initialize(AUDCLNT_SHAREMODE_SHARED, streamflags, p_capture ? REFTIMES_PER_SEC : 0, 0, pwfex, NULL);
 	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
-	event = CreateEvent(NULL, FALSE, FALSE, NULL);
-	ERR_FAIL_COND_V(event == NULL, ERR_CANT_OPEN);
-
-	hr = audio_client->SetEventHandle(event);
+	if (p_capture) {
+		hr = p_device->audio_client->GetService(IID_IAudioCaptureClient, (void **)&p_device->capture_client);
+	} else {
+		hr = p_device->audio_client->GetService(IID_IAudioRenderClient, (void **)&p_device->render_client);
+	}
 	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
-	hr = audio_client->GetService(IID_IAudioRenderClient, (void **)&render_client);
-	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
+	// Free memory
+	CoTaskMemFree(pwfex);
+	SAFE_RELEASE(device)
+
+	return OK;
+}
+
+Error AudioDriverWASAPI::init_render_device(bool reinit) {
+
+	Error err = audio_device_init(&audio_output, false, reinit);
+	if (err != OK)
+		return err;
+
+	switch (audio_output.channels) {
+		case 2: // Stereo
+		case 4: // Surround 3.1
+		case 6: // Surround 5.1
+		case 8: // Surround 7.1
+			channels = audio_output.channels;
+			break;
+
+		default:
+			WARN_PRINTS("WASAPI: Unsupported number of channels: " + itos(audio_output.channels));
+			channels = 2;
+			break;
+	}
 
 	UINT32 max_frames;
-	hr = audio_client->GetBufferSize(&max_frames);
+	HRESULT hr = audio_output.audio_client->GetBufferSize(&max_frames);
 	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
 	// Due to WASAPI Shared Mode we have no control of the buffer size
 	buffer_frames = max_frames;
 
 	// Sample rate is independent of channels (ref: https://stackoverflow.com/questions/11048825/audio-sample-frequency-rely-on-channels)
-	buffer_size = buffer_frames * channels;
-	samples_in.resize(buffer_size);
+	samples_in.resize(buffer_frames * channels);
+
+	input_position = 0;
+	input_size = 0;
 
 	if (OS::get_singleton()->is_stdout_verbose()) {
 		print_line("WASAPI: detected " + itos(channels) + " channels");
@@ -291,41 +331,61 @@ Error AudioDriverWASAPI::init_device(bool reinit) {
 	return OK;
 }
 
-Error AudioDriverWASAPI::finish_device() {
+Error AudioDriverWASAPI::init_capture_device(bool reinit) {
 
-	if (audio_client) {
-		if (active) {
-			audio_client->Stop();
-			active = false;
-		}
+	Error err = audio_device_init(&audio_input, true, reinit);
+	if (err != OK)
+		return err;
 
-		audio_client->Release();
-		audio_client = NULL;
-	}
+	// Get the max frames
+	UINT32 max_frames;
+	HRESULT hr = audio_input.audio_client->GetBufferSize(&max_frames);
+	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
 
-	if (render_client) {
-		render_client->Release();
-		render_client = NULL;
-	}
+	// Set the buffer size
+	input_buffer.resize(max_frames * CAPTURE_BUFFER_CHANNELS);
+	input_position = 0;
+	input_size = 0;
 
-	if (audio_client) {
-		audio_client->Release();
-		audio_client = NULL;
+	return OK;
+}
+
+Error AudioDriverWASAPI::audio_device_finish(AudioDeviceWASAPI *p_device) {
+
+	if (p_device->active) {
+		if (p_device->audio_client) {
+			p_device->audio_client->Stop();
+		}
+
+		p_device->active = false;
 	}
 
+	SAFE_RELEASE(p_device->audio_client)
+	SAFE_RELEASE(p_device->render_client)
+	SAFE_RELEASE(p_device->capture_client)
+
 	return OK;
 }
 
+Error AudioDriverWASAPI::finish_render_device() {
+
+	return audio_device_finish(&audio_output);
+}
+
+Error AudioDriverWASAPI::finish_capture_device() {
+
+	return audio_device_finish(&audio_input);
+}
+
 Error AudioDriverWASAPI::init() {
 
 	mix_rate = GLOBAL_DEF_RST("audio/mix_rate", DEFAULT_MIX_RATE);
 
-	Error err = init_device();
+	Error err = init_render_device();
 	if (err != OK) {
-		ERR_PRINT("WASAPI: init_device error");
+		ERR_PRINT("WASAPI: init_render_device error");
 	}
 
-	active = false;
 	exit_thread = false;
 	thread_exited = false;
 
@@ -345,7 +405,7 @@ AudioDriver::SpeakerMode AudioDriverWASAPI::get_speaker_mode() const {
 	return get_speaker_mode_by_total_channels(channels);
 }
 
-Array AudioDriverWASAPI::get_device_list() {
+Array AudioDriverWASAPI::audio_device_get_list(bool p_capture) {
 
 	Array list;
 	IMMDeviceCollection *devices = NULL;
@@ -358,7 +418,7 @@ Array AudioDriverWASAPI::get_device_list() {
 	HRESULT hr = CoCreateInstance(CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, IID_IMMDeviceEnumerator, (void **)&enumerator);
 	ERR_FAIL_COND_V(hr != S_OK, Array());
 
-	hr = enumerator->EnumAudioEndpoints(eRender, DEVICE_STATE_ACTIVE, &devices);
+	hr = enumerator->EnumAudioEndpoints(p_capture ? eCapture : eRender, DEVICE_STATE_ACTIVE, &devices);
 	ERR_FAIL_COND_V(hr != S_OK, Array());
 
 	UINT count = 0;
@@ -393,21 +453,63 @@ Array AudioDriverWASAPI::get_device_list() {
 	return list;
 }
 
+Array AudioDriverWASAPI::get_device_list() {
+
+	return audio_device_get_list(false);
+}
+
 String AudioDriverWASAPI::get_device() {
 
-	return device_name;
+	lock();
+	String name = audio_output.device_name;
+	unlock();
+
+	return name;
 }
 
 void AudioDriverWASAPI::set_device(String device) {
 
 	lock();
-	new_device = device;
+	audio_output.new_device = device;
 	unlock();
 }
 
-void AudioDriverWASAPI::write_sample(AudioDriverWASAPI *ad, BYTE *buffer, int i, int32_t sample) {
-	if (ad->format_tag == WAVE_FORMAT_PCM) {
-		switch (ad->bits_per_sample) {
+int32_t AudioDriverWASAPI::read_sample(WORD format_tag, int bits_per_sample, BYTE *buffer, int i) {
+	if (format_tag == WAVE_FORMAT_PCM) {
+		int32_t sample = 0;
+		switch (bits_per_sample) {
+			case 8:
+				sample = int32_t(((int8_t *)buffer)[i]) << 24;
+				break;
+
+			case 16:
+				sample = int32_t(((int16_t *)buffer)[i]) << 16;
+				break;
+
+			case 24:
+				sample |= int32_t(((int8_t *)buffer)[i * 3 + 2]) << 24;
+				sample |= int32_t(((int8_t *)buffer)[i * 3 + 1]) << 16;
+				sample |= int32_t(((int8_t *)buffer)[i * 3 + 0]) << 8;
+				break;
+
+			case 32:
+				sample = ((int32_t *)buffer)[i];
+				break;
+		}
+
+		return sample;
+	} else if (format_tag == WAVE_FORMAT_IEEE_FLOAT) {
+		return int32_t(((float *)buffer)[i] * 32768.0) << 16;
+	} else {
+		ERR_PRINT("WASAPI: Unknown format tag");
+	}
+
+	return 0;
+}
+
+void AudioDriverWASAPI::write_sample(WORD format_tag, int bits_per_sample, BYTE *buffer, int i, int32_t sample) {
+	if (format_tag == WAVE_FORMAT_PCM) {
+		switch (bits_per_sample) {
 			case 8:
 				((int8_t *)buffer)[i] = sample >> 24;
 				break;
@@ -426,83 +528,99 @@ void AudioDriverWASAPI::write_sample(AudioDriverWASAPI *ad, BYTE *buffer, int i,
 				((int32_t *)buffer)[i] = sample;
 				break;
 		}
-	} else if (ad->format_tag == WAVE_FORMAT_IEEE_FLOAT) {
+	} else if (format_tag == WAVE_FORMAT_IEEE_FLOAT) {
 		((float *)buffer)[i] = (sample >> 16) / 32768.f;
 	} else {
 		ERR_PRINT("WASAPI: Unknown format tag");
-		ad->exit_thread = true;
 	}
 }
 
 void AudioDriverWASAPI::thread_func(void *p_udata) {
 
 	AudioDriverWASAPI *ad = (AudioDriverWASAPI *)p_udata;
+	uint32_t avail_frames = 0;
+	uint32_t write_ofs = 0;
 
 	while (!ad->exit_thread) {
 
-		ad->lock();
-		ad->start_counting_ticks();
+		uint32_t read_frames = 0;
+		uint32_t written_frames = 0;
 
-		if (ad->active) {
-			ad->audio_server_process(ad->buffer_frames, ad->samples_in.ptrw());
-		} else {
-			for (unsigned int i = 0; i < ad->buffer_size; i++) {
-				ad->samples_in.write[i] = 0;
+		if (avail_frames == 0) {
+			ad->lock();
+			ad->start_counting_ticks();
+
+			if (ad->audio_output.active) {
+				ad->audio_server_process(ad->buffer_frames, ad->samples_in.ptrw());
+			} else {
+				for (unsigned int i = 0; i < ad->samples_in.size(); i++) {
+					ad->samples_in.write[i] = 0;
+				}
 			}
-		}
 
-		ad->stop_counting_ticks();
-		ad->unlock();
+			avail_frames = ad->buffer_frames;
+			write_ofs = 0;
 
-		unsigned int left_frames = ad->buffer_frames;
-		unsigned int buffer_idx = 0;
-		while (left_frames > 0 && ad->audio_client) {
-			WaitForSingleObject(ad->event, 1000);
+			ad->stop_counting_ticks();
+			ad->unlock();
+		}
 
-			ad->lock();
-			ad->start_counting_ticks();
+		ad->lock();
+		ad->start_counting_ticks();
+
+		if (avail_frames > 0 && ad->audio_output.audio_client) {
 
 			UINT32 cur_frames;
 			bool invalidated = false;
-			HRESULT hr = ad->audio_client->GetCurrentPadding(&cur_frames);
+			HRESULT hr = ad->audio_output.audio_client->GetCurrentPadding(&cur_frames);
 			if (hr == S_OK) {
-				// Check how much frames are available on the WASAPI buffer
-				UINT32 avail_frames = ad->buffer_frames - cur_frames;
-				UINT32 write_frames = avail_frames > left_frames ? left_frames : avail_frames;
 
-				BYTE *buffer = NULL;
-				hr = ad->render_client->GetBuffer(write_frames, &buffer);
-				if (hr == S_OK) {
-					// We're using WASAPI Shared Mode so we must convert the buffer
-
-					if (ad->channels == ad->wasapi_channels) {
-						for (unsigned int i = 0; i < write_frames * ad->channels; i++) {
-							ad->write_sample(ad, buffer, i, ad->samples_in[buffer_idx++]);
-						}
-					} else {
-						for (unsigned int i = 0; i < write_frames; i++) {
-							for (unsigned int j = 0; j < MIN(ad->channels, ad->wasapi_channels); j++) {
-								ad->write_sample(ad, buffer, i * ad->wasapi_channels + j, ad->samples_in[buffer_idx++]);
+				// Check how much frames are available on the WASAPI buffer
+				UINT32 write_frames = MIN(ad->buffer_frames - cur_frames, avail_frames);
+				if (write_frames > 0) {
+					BYTE *buffer = NULL;
+					hr = ad->audio_output.render_client->GetBuffer(write_frames, &buffer);
+					if (hr == S_OK) {
+
+						// We're using WASAPI Shared Mode so we must convert the buffer
+						if (ad->channels == ad->audio_output.channels) {
+							for (unsigned int i = 0; i < write_frames * ad->channels; i++) {
+								ad->write_sample(ad->audio_output.format_tag, ad->audio_output.bits_per_sample, buffer, i, ad->samples_in.write[write_ofs++]);
 							}
-							if (ad->wasapi_channels > ad->channels) {
-								for (unsigned int j = ad->channels; j < ad->wasapi_channels; j++) {
-									ad->write_sample(ad, buffer, i * ad->wasapi_channels + j, 0);
+						} else {
+							for (unsigned int i = 0; i < write_frames; i++) {
+								for (unsigned int j = 0; j < MIN(ad->channels, ad->audio_output.channels); j++) {
+									ad->write_sample(ad->audio_output.format_tag, ad->audio_output.bits_per_sample, buffer, i * ad->audio_output.channels + j, ad->samples_in.write[write_ofs++]);
+								}
+								if (ad->audio_output.channels > ad->channels) {
+									for (unsigned int j = ad->channels; j < ad->audio_output.channels; j++) {
+										ad->write_sample(ad->audio_output.format_tag, ad->audio_output.bits_per_sample, buffer, i * ad->audio_output.channels + j, 0);
+									}
 								}
 							}
 						}
-					}
 
-					hr = ad->render_client->ReleaseBuffer(write_frames, 0);
-					if (hr != S_OK) {
-						ERR_PRINT("WASAPI: Release buffer error");
-					}
+						hr = ad->audio_output.render_client->ReleaseBuffer(write_frames, 0);
+						if (hr != S_OK) {
+							ERR_PRINT("WASAPI: Release buffer error");
+						}
 
-					left_frames -= write_frames;
-				} else if (hr == AUDCLNT_E_DEVICE_INVALIDATED) {
-					invalidated = true;
-				} else {
-					ERR_PRINT("WASAPI: Get buffer error");
-					ad->exit_thread = true;
+						avail_frames -= write_frames;
+						written_frames += write_frames;
+					} else if (hr == AUDCLNT_E_DEVICE_INVALIDATED) {
+						// Device is not valid anymore, reopen it
+
+						Error err = ad->finish_render_device();
+						if (err != OK) {
+							ERR_PRINT("WASAPI: finish_render_device error");
+						} else {
+							// We reopened the device and samples_in may have resized, so invalidate the current avail_frames
+							avail_frames = 0;
+						}
+					} else {
+						ERR_PRINT("WASAPI: Get buffer error");
+						ad->exit_thread = true;
+					}
 				}
 			} else if (hr == AUDCLNT_E_DEVICE_INVALIDATED) {
 				invalidated = true;
@@ -514,47 +632,117 @@ void AudioDriverWASAPI::thread_func(void *p_udata) {
 				// Device is not valid anymore
 				WARN_PRINT("WASAPI: Current device invalidated, closing device");
 
-				Error err = ad->finish_device();
+				Error err = ad->finish_render_device();
 				if (err != OK) {
-					ERR_PRINT("WASAPI: finish_device error");
+					ERR_PRINT("WASAPI: finish_render_device error");
 				}
 			}
-
-			ad->stop_counting_ticks();
-			ad->unlock();
 		}
 
-		ad->lock();
-		ad->start_counting_ticks();
-
 		// If we're using the Default device and it changed finish it so we'll re-init the device
-		if (ad->device_name == "Default" && default_device_changed) {
-			Error err = ad->finish_device();
+		if (ad->audio_output.device_name == "Default" && default_render_device_changed) {
+			Error err = ad->finish_render_device();
 			if (err != OK) {
-				ERR_PRINT("WASAPI: finish_device error");
+				ERR_PRINT("WASAPI: finish_render_device error");
 			}
 
-			default_device_changed = false;
+			default_render_device_changed = false;
 		}
 
 		// User selected a new device, finish the current one so we'll init the new device
-		if (ad->device_name != ad->new_device) {
-			ad->device_name = ad->new_device;
-			Error err = ad->finish_device();
+		if (ad->audio_output.device_name != ad->audio_output.new_device) {
+			ad->audio_output.device_name = ad->audio_output.new_device;
+			Error err = ad->finish_render_device();
 			if (err != OK) {
-				ERR_PRINT("WASAPI: finish_device error");
+				ERR_PRINT("WASAPI: finish_render_device error");
 			}
 		}
 
-		if (!ad->audio_client) {
-			Error err = ad->init_device(true);
+		if (!ad->audio_output.audio_client) {
+			Error err = ad->init_render_device(true);
 			if (err == OK) {
 				ad->start();
 			}
 		}
 
+		if (ad->audio_input.active) {
+			UINT32 packet_length = 0;
+			BYTE *data;
+			UINT32 num_frames_available;
+			DWORD flags;
+
+			HRESULT hr = ad->audio_input.capture_client->GetNextPacketSize(&packet_length);
+			if (hr == S_OK) {
+				while (packet_length != 0) {
+					hr = ad->audio_input.capture_client->GetBuffer(&data, &num_frames_available, &flags, NULL, NULL);
+					ERR_BREAK(hr != S_OK);
+
+					// fixme: Only works for floating point atm
+					for (int j = 0; j < num_frames_available; j++) {
+						int32_t l, r;
+
+						if (flags & AUDCLNT_BUFFERFLAGS_SILENT) {
+							l = r = 0;
+						} else {
+							if (ad->audio_input.channels == 2) {
+								l = read_sample(ad->audio_input.format_tag, ad->audio_input.bits_per_sample, data, j * 2);
+								r = read_sample(ad->audio_input.format_tag, ad->audio_input.bits_per_sample, data, j * 2 + 1);
+							} else if (ad->audio_input.channels == 1) {
+								l = r = read_sample(ad->audio_input.format_tag, ad->audio_input.bits_per_sample, data, j);
+							} else {
+								l = r = 0;
+								ERR_PRINT("WASAPI: unsupported channel count in microphone!");
+							}
+						}
+
+						ad->input_buffer_write(l);
+						ad->input_buffer_write(r);
+					}
+
+					read_frames += num_frames_available;
+
+					hr = ad->audio_input.capture_client->ReleaseBuffer(num_frames_available);
+					ERR_BREAK(hr != S_OK);
+
+					hr = ad->audio_input.capture_client->GetNextPacketSize(&packet_length);
+					ERR_BREAK(hr != S_OK);
+				}
+			}
+
+			// If we're using the Default device and it changed finish it so we'll re-init the device
+			if (ad->audio_input.device_name == "Default" && default_capture_device_changed) {
+				Error err = ad->finish_capture_device();
+				if (err != OK) {
+					ERR_PRINT("WASAPI: finish_capture_device error");
+				}
+
+				default_capture_device_changed = false;
+			}
+
+			// User selected a new device, finish the current one so we'll init the new device
+			if (ad->audio_input.device_name != ad->audio_input.new_device) {
+				ad->audio_input.device_name = ad->audio_input.new_device;
+				Error err = ad->finish_capture_device();
+				if (err != OK) {
+					ERR_PRINT("WASAPI: finish_capture_device error");
+				}
+			}
+
+			if (!ad->audio_input.audio_client) {
+				Error err = ad->init_capture_device(true);
+				if (err == OK) {
+					ad->capture_start();
+				}
+			}
+		}
+
 		ad->stop_counting_ticks();
 		ad->unlock();
+
+		// Let the thread rest a while if we haven't read or write anything
+		if (written_frames == 0 && read_frames == 0) {
+			OS::get_singleton()->delay_usec(1000);
+		}
 	}
 
 	ad->thread_exited = true;
@@ -562,12 +750,12 @@ void AudioDriverWASAPI::thread_func(void *p_udata) {
 
 void AudioDriverWASAPI::start() {
 
-	if (audio_client) {
-		HRESULT hr = audio_client->Start();
+	if (audio_output.audio_client) {
+		HRESULT hr = audio_output.audio_client->Start();
 		if (hr != S_OK) {
 			ERR_PRINT("WASAPI: Start failed");
 		} else {
-			active = true;
+			audio_output.active = true;
 		}
 	}
 }
@@ -594,7 +782,8 @@ void AudioDriverWASAPI::finish() {
 		thread = NULL;
 	}
 
-	finish_device();
+	finish_capture_device();
+	finish_render_device();
 
 	if (mutex) {
 		memdelete(mutex);
@@ -602,30 +791,70 @@ void AudioDriverWASAPI::finish() {
 	}
 }
 
+Error AudioDriverWASAPI::capture_start() {
+
+	Error err = init_capture_device();
+	if (err != OK) {
+		ERR_PRINT("WASAPI: init_capture_device error");
+		return err;
+	}
+
+	if (audio_input.active == false) {
+		audio_input.audio_client->Start();
+		audio_input.active = true;
+
+		return OK;
+	}
+
+	return FAILED;
+}
+
+Error AudioDriverWASAPI::capture_stop() {
+
+	if (audio_input.active == true) {
+		audio_input.audio_client->Stop();
+		audio_input.active = false;
+
+		return OK;
+	}
+
+	return FAILED;
+}
+
+void AudioDriverWASAPI::capture_set_device(const String &p_name) {
+
+	lock();
+	audio_input.new_device = p_name;
+	unlock();
+}
+
+Array AudioDriverWASAPI::capture_get_device_list() {
+
+	return audio_device_get_list(true);
+}
+
+String AudioDriverWASAPI::capture_get_device() {
+
+	lock();
+	String name = audio_input.device_name;
+	unlock();
+
+	return name;
+}
+
 AudioDriverWASAPI::AudioDriverWASAPI() {
 
-	audio_client = NULL;
-	render_client = NULL;
 	mutex = NULL;
 	thread = NULL;
 
-	format_tag = 0;
-	bits_per_sample = 0;
-
 	samples_in.clear();
 
-	buffer_size = 0;
 	channels = 0;
-	wasapi_channels = 0;
 	mix_rate = 0;
 	buffer_frames = 0;
 
 	thread_exited = false;
 	exit_thread = false;
-	active = false;
-
-	device_name = "Default";
-	new_device = "Default";
 }
 
 #endif

drivers/wasapi/audio_driver_wasapi.h

@@ -43,35 +43,63 @@
 
 class AudioDriverWASAPI : public AudioDriver {
 
-	HANDLE event;
-	IAudioClient *audio_client;
-	IAudioRenderClient *render_client;
+	class AudioDeviceWASAPI {
+	public:
+		IAudioClient *audio_client;
+		IAudioRenderClient *render_client;
+		IAudioCaptureClient *capture_client;
+		bool active;
+
+		WORD format_tag;
+		WORD bits_per_sample;
+		unsigned int channels;
+		unsigned int frame_size;
+
+		String device_name;
+		String new_device;
+
+		AudioDeviceWASAPI() {
+			audio_client = NULL;
+			render_client = NULL;
+			capture_client = NULL;
+			active = false;
+			format_tag = 0;
+			bits_per_sample = 0;
+			channels = 0;
+			frame_size = 0;
+			device_name = "Default";
+			new_device = "Default";
+		}
+	};
+
+	AudioDeviceWASAPI audio_input;
+	AudioDeviceWASAPI audio_output;
+
 	Mutex *mutex;
 	Thread *thread;
 
-	String device_name;
-	String new_device;
-
-	WORD format_tag;
-	WORD bits_per_sample;
-
 	Vector<int32_t> samples_in;
 
-	unsigned int buffer_size;
 	unsigned int channels;
-	unsigned int wasapi_channels;
 	int mix_rate;
 	int buffer_frames;
 
 	bool thread_exited;
 	mutable bool exit_thread;
-	bool active;
 
-	_FORCE_INLINE_ void write_sample(AudioDriverWASAPI *ad, BYTE *buffer, int i, int32_t sample);
+	static _FORCE_INLINE_ void write_sample(WORD format_tag, int bits_per_sample, BYTE *buffer, int i, int32_t sample);
+	static _FORCE_INLINE_ int32_t read_sample(WORD format_tag, int bits_per_sample, BYTE *buffer, int i);
 	static void thread_func(void *p_udata);
 
-	Error init_device(bool reinit = false);
-	Error finish_device();
+	Error init_render_device(bool reinit = false);
+	Error init_capture_device(bool reinit = false);
+
+	Error finish_render_device();
+	Error finish_capture_device();
+
+	Error audio_device_init(AudioDeviceWASAPI *p_device, bool p_capture, bool reinit);
+	Error audio_device_finish(AudioDeviceWASAPI *p_device);
+	Array audio_device_get_list(bool p_capture);
 
 public:
 	virtual const char *get_name() const {
@@ -89,6 +117,12 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
+	virtual Error capture_start();
+	virtual Error capture_stop();
+	virtual Array capture_get_device_list();
+	virtual void capture_set_device(const String &p_name);
+	virtual String capture_get_device();
+
 	AudioDriverWASAPI();
 };
 

servers/audio/audio_stream.cpp

@@ -29,6 +29,7 @@
 /*************************************************************************/
 
 #include "audio_stream.h"
+#include "os/os.h"
 
 //////////////////////////////
 
@@ -99,6 +100,119 @@ void AudioStream::_bind_methods() {
 
 ////////////////////////////////
 
+Ref<AudioStreamPlayback> AudioStreamMicrophone::instance_playback() {
+	Ref<AudioStreamPlaybackMicrophone> playback;
+	playback.instance();
+
+	playbacks.insert(playback.ptr());
+
+	playback->microphone = Ref<AudioStreamMicrophone>((AudioStreamMicrophone *)this);
+	playback->active = false;
+
+	return playback;
+}
+
+String AudioStreamMicrophone::get_stream_name() const {
+
+	//if (audio_stream.is_valid()) {
+	//return "Random: " + audio_stream->get_name();
+	//}
+	return "Microphone";
+}
+
+float AudioStreamMicrophone::get_length() const {
+	return 0;
+}
+
+void AudioStreamMicrophone::_bind_methods() {
+}
+
+AudioStreamMicrophone::AudioStreamMicrophone() {
+}
+
+void AudioStreamPlaybackMicrophone::_mix_internal(AudioFrame *p_buffer, int p_frames) {
+
+	AudioDriver::get_singleton()->lock();
+
+	Vector<int32_t> buf = AudioDriver::get_singleton()->get_input_buffer();
+	unsigned int input_size = AudioDriver::get_singleton()->get_input_size();
+
+	// p_frames is multipled by two since an AudioFrame is stereo
+	if ((p_frames + MICROPHONE_PLAYBACK_DELAY * 2) > input_size) {
+		for (int i = 0; i < p_frames; i++) {
+			p_buffer[i] = AudioFrame(0.0f, 0.0f);
+		}
+		input_ofs = 0;
+	} else {
+		for (int i = 0; i < p_frames; i++) {
+			if (input_size >= input_ofs) {
+				float l = (buf[input_ofs++] >> 16) / 32768.f;
+				if (input_ofs >= buf.size()) {
+					input_ofs = 0;
+				}
+				float r = (buf[input_ofs++] >> 16) / 32768.f;
+				if (input_ofs >= buf.size()) {
+					input_ofs = 0;
+				}
+
+				p_buffer[i] = AudioFrame(l, r);
+			} else {
+				p_buffer[i] = AudioFrame(0.0f, 0.0f);
+			}
+		}
+	}
+
+	AudioDriver::get_singleton()->unlock();
+}
+
+void AudioStreamPlaybackMicrophone::mix(AudioFrame *p_buffer, float p_rate_scale, int p_frames) {
+	AudioStreamPlaybackResampled::mix(p_buffer, p_rate_scale, p_frames);
+}
+
+float AudioStreamPlaybackMicrophone::get_stream_sampling_rate() {
+	return AudioDriver::get_singleton()->get_mix_rate();
+}
+
+void AudioStreamPlaybackMicrophone::start(float p_from_pos) {
+	input_ofs = 0;
+
+	AudioDriver::get_singleton()->capture_start();
+
+	active = true;
+	_begin_resample();
+}
+
+void AudioStreamPlaybackMicrophone::stop() {
+	AudioDriver::get_singleton()->capture_stop();
+	active = false;
+}
+
+bool AudioStreamPlaybackMicrophone::is_playing() const {
+	return active;
+}
+
+int AudioStreamPlaybackMicrophone::get_loop_count() const {
+	return 0;
+}
+
+float AudioStreamPlaybackMicrophone::get_playback_position() const {
+	return 0;
+}
+
+void AudioStreamPlaybackMicrophone::seek(float p_time) {
+	return; // Can't seek a microphone input
+}
+
+AudioStreamPlaybackMicrophone::~AudioStreamPlaybackMicrophone() {
+	microphone->playbacks.erase(this);
+	stop();
+}
+
+AudioStreamPlaybackMicrophone::AudioStreamPlaybackMicrophone() {
+}
+
+////////////////////////////////
+
 void AudioStreamRandomPitch::set_audio_stream(const Ref<AudioStream> &p_audio_stream) {
 
 	audio_stream = p_audio_stream;

servers/audio/audio_stream.h

@@ -94,6 +94,63 @@ public:
 	virtual float get_length() const = 0; //if supported, otherwise return 0
 };
 
+// Microphone
+
+class AudioStreamPlaybackMicrophone;
+
+class AudioStreamMicrophone : public AudioStream {
+
+	GDCLASS(AudioStreamMicrophone, AudioStream)
+	friend class AudioStreamPlaybackMicrophone;
+
+	Set<AudioStreamPlaybackMicrophone *> playbacks;
+
+protected:
+	static void _bind_methods();
+
+public:
+	virtual Ref<AudioStreamPlayback> instance_playback();
+	virtual String get_stream_name() const;
+
+	virtual float get_length() const; //if supported, otherwise return 0
+
+	AudioStreamMicrophone();
+};
+
+class AudioStreamPlaybackMicrophone : public AudioStreamPlaybackResampled {
+
+	GDCLASS(AudioStreamPlaybackMicrophone, AudioStreamPlayback)
+	friend class AudioStreamMicrophone;
+
+	const int MICROPHONE_PLAYBACK_DELAY = 256;
+
+	bool active;
+	unsigned int input_ofs;
+
+	Ref<AudioStreamMicrophone> microphone;
+
+protected:
+	virtual void _mix_internal(AudioFrame *p_buffer, int p_frames);
+	virtual float get_stream_sampling_rate();
+
+public:
+	virtual void mix(AudioFrame *p_buffer, float p_rate_scale, int p_frames);
+
+	virtual void start(float p_from_pos = 0.0);
+	virtual void stop();
+	virtual bool is_playing() const;
+
+	virtual int get_loop_count() const; //times it looped
+
+	virtual float get_playback_position() const;
+	virtual void seek(float p_time);
+
+	~AudioStreamPlaybackMicrophone();
+	AudioStreamPlaybackMicrophone();
+};
+
+//
+
 class AudioStreamPlaybackRandomPitch;
 
 class AudioStreamRandomPitch : public AudioStream {

servers/audio_server.cpp

@@ -33,6 +33,7 @@
 #include "os/file_access.h"
 #include "os/os.h"
 #include "project_settings.h"
+#include "scene/resources/audio_stream_sample.h"
 #include "servers/audio/audio_driver_dummy.h"
 #include "servers/audio/effects/audio_effect_compressor.h"
 #ifdef TOOLS_ENABLED
@@ -79,6 +80,17 @@ double AudioDriver::get_mix_time() const {
 	return total;
 }
 
+void AudioDriver::input_buffer_write(int32_t sample) {
+
+	input_buffer.write[input_position++] = sample;
+	if (input_position >= input_buffer.size()) {
+		input_position = 0;
+	}
+	if (input_size < input_buffer.size()) {
+		input_size++;
+	}
+}
+
 AudioDriver::SpeakerMode AudioDriver::get_speaker_mode_by_total_channels(int p_channels) const {
 	switch (p_channels) {
 		case 4: return SPEAKER_SURROUND_31;
@@ -113,6 +125,14 @@ String AudioDriver::get_device() {
 	return "Default";
 }
 
+Array AudioDriver::capture_get_device_list() {
+	Array list;
+
+	list.push_back("Default");
+
+	return list;
+}
+
 AudioDriver::AudioDriver() {
 
 	_last_mix_time = 0;
@@ -1201,6 +1221,21 @@ void AudioServer::set_device(String device) {
 	AudioDriver::get_singleton()->set_device(device);
 }
 
+Array AudioServer::capture_get_device_list() {
+
+	return AudioDriver::get_singleton()->capture_get_device_list();
+}
+
+String AudioServer::capture_get_device() {
+
+	return AudioDriver::get_singleton()->capture_get_device();
+}
+
+void AudioServer::capture_set_device(const String &p_name) {
+
+	AudioDriver::get_singleton()->capture_set_device(p_name);
+}
+
 void AudioServer::_bind_methods() {
 
 	ClassDB::bind_method(D_METHOD("set_bus_count", "amount"), &AudioServer::set_bus_count);
@@ -1251,6 +1286,10 @@ void AudioServer::_bind_methods() {
 	ClassDB::bind_method(D_METHOD("get_device"), &AudioServer::get_device);
 	ClassDB::bind_method(D_METHOD("set_device"), &AudioServer::set_device);
 
+	ClassDB::bind_method(D_METHOD("capture_get_device_list"), &AudioServer::capture_get_device_list);
+	ClassDB::bind_method(D_METHOD("capture_get_device"), &AudioServer::capture_get_device);
+	ClassDB::bind_method(D_METHOD("capture_set_device"), &AudioServer::capture_set_device);
+
 	ClassDB::bind_method(D_METHOD("set_bus_layout", "bus_layout"), &AudioServer::set_bus_layout);
 	ClassDB::bind_method(D_METHOD("generate_bus_layout"), &AudioServer::generate_bus_layout);
 

servers/audio_server.h

@@ -38,6 +38,8 @@
 #include "variant.h"
 
 class AudioDriverDummy;
+class AudioStream;
+class AudioStreamSample;
 
 class AudioDriver {
 
@@ -51,8 +53,13 @@ class AudioDriver {
 #endif
 
 protected:
+	Vector<int32_t> input_buffer;
+	unsigned int input_position;
+	unsigned int input_size;
+
 	void audio_server_process(int p_frames, int32_t *p_buffer, bool p_update_mix_time = true);
 	void update_mix_time(int p_frames);
+	void input_buffer_write(int32_t sample);
 
 #ifdef DEBUG_ENABLED
 	_FORCE_INLINE_ void start_counting_ticks() { prof_ticks = OS::get_singleton()->get_ticks_usec(); }
@@ -91,11 +98,21 @@ public:
 	virtual void unlock() = 0;
 	virtual void finish() = 0;
 
+	virtual Error capture_start() { return FAILED; }
+	virtual Error capture_stop() { return FAILED; }
+	virtual void capture_set_device(const String &p_name) {}
+	virtual String capture_get_device() { return "Default"; }
+	virtual Array capture_get_device_list(); // TODO: convert this and get_device_list to PoolStringArray
+
 	virtual float get_latency() { return 0; }
 
 	SpeakerMode get_speaker_mode_by_total_channels(int p_channels) const;
 	int get_total_channels_by_speaker_mode(SpeakerMode) const;
 
+	Vector<int32_t> get_input_buffer() { return input_buffer; }
+	unsigned int get_input_position() { return input_position; }
+	unsigned int get_input_size() { return input_size; }
+
 #ifdef DEBUG_ENABLED
 	uint64_t get_profiling_time() const { return prof_time; }
 	void reset_profiling_time() { prof_time = 0; }
@@ -222,6 +239,18 @@ private:
 
 	void _mix_step();
 
+#if 0
+	struct AudioInBlock {
+
+		Ref<AudioStreamSample> audio_stream;
+		int current_position;
+		bool loops;
+	};
+
+	Map<StringName, AudioInBlock *> audio_in_block_map;
+	Vector<AudioInBlock *> audio_in_blocks;
+#endif
+
 	struct CallbackItem {
 
 		AudioCallback callback;
@@ -335,8 +364,11 @@ public:
 	String get_device();
 	void set_device(String device);
 
-	float get_output_latency() { return output_latency; }
+	Array capture_get_device_list();
+	String capture_get_device();
+	void capture_set_device(const String &p_name);
 
+	float get_output_latency() { return output_latency; }
 	AudioServer();
 	virtual ~AudioServer();
 };

servers/register_server_types.cpp

@@ -104,6 +104,7 @@ void register_server_types() {
 
 	ClassDB::register_virtual_class<AudioStream>();
 	ClassDB::register_virtual_class<AudioStreamPlayback>();
+	ClassDB::register_class<AudioStreamMicrophone>();
 	ClassDB::register_class<AudioStreamRandomPitch>();
 	ClassDB::register_virtual_class<AudioEffect>();
 	ClassDB::register_class<AudioEffectEQ>();