Modified Microphone implementation to handle only one device at a time (WIP)

22 files changed, 531 insertions, 993 deletions

diff --git a/drivers/alsa/audio_driver_alsa.cpp b/drivers/alsa/audio_driver_alsa.cpp
index 29b1207927..a44a11a46d 100644
--- a/drivers/alsa/audio_driver_alsa.cpp
+++ b/drivers/alsa/audio_driver_alsa.cpp
@@ -339,28 +339,6 @@ void AudioDriverALSA::finish() {
 	finish_device();
 }
 
-bool AudioDriverALSA::capture_device_start(StringName p_name) {
-
-	return false;
-}
-
-bool AudioDriverALSA::capture_device_stop(StringName p_name) {
-
-	return false;
-}
-
-PoolStringArray AudioDriverALSA::capture_device_get_names() {
-
-	PoolStringArray names;
-
-	return names;
-}
-
-StringName AudioDriverALSA::capture_device_get_default_name() {
-
-	return "";
-}
-
 AudioDriverALSA::AudioDriverALSA() {
 
 	mutex = NULL;
diff --git a/drivers/alsa/audio_driver_alsa.h b/drivers/alsa/audio_driver_alsa.h
index 59bc7c4968..e2a2325cf3 100644
--- a/drivers/alsa/audio_driver_alsa.h
+++ b/drivers/alsa/audio_driver_alsa.h
@@ -83,11 +83,6 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	AudioDriverALSA();
 	~AudioDriverALSA();
 };
diff --git a/drivers/coreaudio/audio_driver_coreaudio.cpp b/drivers/coreaudio/audio_driver_coreaudio.cpp
index e13b81ddab..261ba7809c 100644
--- a/drivers/coreaudio/audio_driver_coreaudio.cpp
+++ b/drivers/coreaudio/audio_driver_coreaudio.cpp
@@ -35,6 +35,7 @@
 #include "os/os.h"
 
 #define kOutputBus 0
+#define kInputBus 1
 
 #ifdef OSX_ENABLED
 OSStatus AudioDriverCoreAudio::output_device_address_cb(AudioObjectID inObjectID,
@@ -117,6 +118,11 @@ Error AudioDriverCoreAudio::init() {
 	result = AudioUnitSetProperty(audio_unit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Input, kOutputBus, &strdesc, sizeof(strdesc));
 	ERR_FAIL_COND_V(result != noErr, FAILED);
 
+	strdesc.mChannelsPerFrame = 2;
+
+	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 +132,14 @@ 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);
+	audio_input_buffer.resize(buffer_size * 8);
+	for (int i = 0; i < audio_input_buffer.size(); i++) {
+		audio_input_buffer.write[i] = 0;
+	}
+	audio_input_position = 0;
 
 	if (OS::get_singleton()->is_stdout_verbose()) {
 		print_line("CoreAudio: detected " + itos(channels) + " channels");
@@ -141,6 +153,12 @@ 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 +210,42 @@ 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 = 2;
+	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 * 2; i++) {
+			ad->audio_input_buffer.write[ad->audio_input_position++] = ad->input_buf[i] << 16;
+			if (ad->audio_input_position >= ad->audio_input_buffer.size()) {
+				ad->audio_input_position = 0;
+			}
+		}
+	} 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);
@@ -434,26 +488,22 @@ void AudioDriverCoreAudio::finish() {
 	}
 };
 
-bool AudioDriverCoreAudio::capture_device_start(StringName p_name) {
+Error AudioDriverCoreAudio::capture_start() {
 
-	return false;
-}
-
-bool AudioDriverCoreAudio::capture_device_stop(StringName p_name) {
+	UInt32 flag = 1;
+	OSStatus result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Input, kInputBus, &flag, sizeof(flag));
+	ERR_FAIL_COND_V(result != noErr, FAILED);
 
-	return false;
+	return OK;
 }
 
-PoolStringArray AudioDriverCoreAudio::capture_device_get_names() {
-
-	PoolStringArray names;
+Error AudioDriverCoreAudio::capture_stop() {
 
-	return names;
-}
-
-StringName AudioDriverCoreAudio::capture_device_get_default_name() {
+	UInt32 flag = 0;
+	OSStatus result = AudioUnitSetProperty(audio_unit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Input, kInputBus, &flag, sizeof(flag));
+	ERR_FAIL_COND_V(result != noErr, FAILED);
 
-	return "";
+	return OK;
 }
 
 AudioDriverCoreAudio::AudioDriverCoreAudio() {
@@ -463,7 +513,6 @@ AudioDriverCoreAudio::AudioDriverCoreAudio() {
 	mix_rate = 0;
 	channels = 2;
 
-	buffer_size = 0;
 	buffer_frames = 0;
 
 	samples_in.clear();
diff --git a/drivers/coreaudio/audio_driver_coreaudio.h b/drivers/coreaudio/audio_driver_coreaudio.h
index 0d3efca28d..7629e56686 100644
--- a/drivers/coreaudio/audio_driver_coreaudio.h
+++ b/drivers/coreaudio/audio_driver_coreaudio.h
@@ -52,9 +52,9 @@ class AudioDriverCoreAudio : public AudioDriver {
 	int mix_rate;
 	unsigned int channels;
 	unsigned int buffer_frames;
-	unsigned int buffer_size;
 
 	Vector<int32_t> samples_in;
+	Vector<int16_t> input_buf;
 
 #ifdef OSX_ENABLED
 	static OSStatus output_device_address_cb(AudioObjectID inObjectID,
@@ -68,6 +68,12 @@ 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";
@@ -86,10 +92,8 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
+	virtual Error capture_start();
+	virtual Error capture_stop();
 
 	bool try_lock();
 	void stop();
diff --git a/drivers/pulseaudio/audio_driver_pulseaudio.cpp b/drivers/pulseaudio/audio_driver_pulseaudio.cpp
index 49d3ab2070..2b1a336630 100644
--- a/drivers/pulseaudio/audio_driver_pulseaudio.cpp
+++ b/drivers/pulseaudio/audio_driver_pulseaudio.cpp
@@ -287,74 +287,71 @@ 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();
+		size_t read_bytes = 0;
+		size_t written_bytes = 0;
 
-		if (!ad->active) {
-			for (unsigned int i = 0; i < ad->pa_buffer_size; i++) {
-				ad->samples_out.write[i] = 0;
-			}
+		if (avail_bytes == 0) {
+			ad->lock();
+			ad->start_counting_ticks();
 
-		} else {
-			ad->audio_server_process(ad->buffer_frames, ad->samples_in.ptrw());
-
-			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 +376,41 @@ 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->audio_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--) {
+						ad->audio_input_buffer.write[ad->audio_input_position++] = int32_t(*srcptr++) << 16;
+						if (ad->audio_input_position >= ad->audio_input_buffer.size()) {
+							ad->audio_input_position = 0;
+						}
+					}
+
+					read_bytes += bytes;
+					ret = pa_stream_drop(ad->pa_rec_str);
+					if (ret != 0) {
+						ERR_PRINT("pa_stream_drop error");
+					}
+				}
+			}
+		}
+
 		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;
@@ -510,26 +540,60 @@ void AudioDriverPulseAudio::finish() {
 	thread = NULL;
 }
 
-bool AudioDriverPulseAudio::capture_device_start(StringName p_name) {
+Error AudioDriverPulseAudio::capture_start() {
 
-	return false;
-}
+	Error err = OK;
 
-bool AudioDriverPulseAudio::capture_device_stop(StringName p_name) {
+	lock();
 
-	return false;
-}
+	pa_sample_spec spec;
 
-PoolStringArray AudioDriverPulseAudio::capture_device_get_names() {
+	spec.format = PA_SAMPLE_S16LE;
+	spec.channels = 2;
+	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;
 
-	PoolStringArray names;
+	pa_buffer_attr attr;
+	attr.fragsize = buffer_size * sizeof(int16_t);
+
+	pa_channel_map pa_rec_map;
+	pa_channel_map_init_stereo(&pa_rec_map);
 
-	return names;
+	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);
+	}
+
+	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, NULL, &attr, flags);
+	if (error_code < 0) {
+		ERR_PRINTS("PulseAudio: pa_stream_connect_record error: " + String(pa_strerror(error_code)));
+		err = ERR_CANT_OPEN;
+	}
+
+	audio_input_buffer.resize(input_buffer_frames * 8);
+	for (int i = 0; i < audio_input_buffer.size(); i++) {
+		audio_input_buffer.write[i] = 0;
+	}
+	audio_input_position = 0;
+
+	unlock();
+
+	return err;
 }
 
-StringName AudioDriverPulseAudio::capture_device_get_default_name() {
+Error AudioDriverPulseAudio::capture_stop() {
+	if (pa_rec_str) {
+		pa_stream_disconnect(pa_rec_str);
+		pa_stream_unref(pa_rec_str);
+		pa_rec_str = NULL;
+	}
 
-	return "";
+	return OK;
 }
 
 AudioDriverPulseAudio::AudioDriverPulseAudio() {
@@ -537,6 +601,7 @@ AudioDriverPulseAudio::AudioDriverPulseAudio() {
 	pa_ml = NULL;
 	pa_ctx = NULL;
 	pa_str = NULL;
+	pa_rec_str = NULL;
 
 	mutex = NULL;
 	thread = NULL;
@@ -550,6 +615,7 @@ AudioDriverPulseAudio::AudioDriverPulseAudio() {
 
 	mix_rate = 0;
 	buffer_frames = 0;
+	input_buffer_frames = 0;
 	pa_buffer_size = 0;
 	channels = 0;
 	pa_ready = 0;
diff --git a/drivers/pulseaudio/audio_driver_pulseaudio.h b/drivers/pulseaudio/audio_driver_pulseaudio.h
index bfbd340e26..67b99133fb 100644
--- a/drivers/pulseaudio/audio_driver_pulseaudio.h
+++ b/drivers/pulseaudio/audio_driver_pulseaudio.h
@@ -47,6 +47,7 @@ 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;
 
 	String device_name;
@@ -58,6 +59,7 @@ class AudioDriverPulseAudio : public AudioDriver {
 
 	unsigned int mix_rate;
 	unsigned int buffer_frames;
+	unsigned int input_buffer_frames;
 	unsigned int pa_buffer_size;
 	int channels;
 	int pa_ready;
@@ -98,13 +100,11 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	virtual float get_latency();
 
+	virtual Error capture_start();
+	virtual Error capture_stop();
+
 	AudioDriverPulseAudio();
 	~AudioDriverPulseAudio();
 };
diff --git a/drivers/rtaudio/audio_driver_rtaudio.cpp b/drivers/rtaudio/audio_driver_rtaudio.cpp
index 0fad96f133..365788e192 100644
--- a/drivers/rtaudio/audio_driver_rtaudio.cpp
+++ b/drivers/rtaudio/audio_driver_rtaudio.cpp
@@ -194,28 +194,6 @@ void AudioDriverRtAudio::finish() {
 	}
 }
 
-bool AudioDriverRtAudio::capture_device_start(StringName p_name) {
-
-	return false;
-}
-
-bool AudioDriverRtAudio::capture_device_stop(StringName p_name) {
-
-	return false;
-}
-
-PoolStringArray AudioDriverRtAudio::capture_device_get_names() {
-
-	PoolStringArray names;
-
-	return names;
-}
-
-StringName AudioDriverRtAudio::capture_device_get_default_name() {
-
-	return "";
-}
-
 AudioDriverRtAudio::AudioDriverRtAudio() {
 
 	active = false;
diff --git a/drivers/rtaudio/audio_driver_rtaudio.h b/drivers/rtaudio/audio_driver_rtaudio.h
index 789a21ee32..2a64652d5f 100644
--- a/drivers/rtaudio/audio_driver_rtaudio.h
+++ b/drivers/rtaudio/audio_driver_rtaudio.h
@@ -58,11 +58,6 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	AudioDriverRtAudio();
 };
 
diff --git a/drivers/wasapi/audio_driver_wasapi.cpp b/drivers/wasapi/audio_driver_wasapi.cpp
index 8fe83a3be5..f8d7516f1f 100644
--- a/drivers/wasapi/audio_driver_wasapi.cpp
+++ b/drivers/wasapi/audio_driver_wasapi.cpp
@@ -142,7 +142,7 @@ public:
 
 static CMMNotificationClient notif_client;
 
-Error AudioDriverWASAPI::init_render_device(bool reinit) {
+Error AudioDriverWASAPI::audio_device_init(AudioDeviceWASAPI *p_device, bool p_capture, bool reinit) {
 
 	WAVEFORMATEX *pwfex;
 	IMMDeviceEnumerator *enumerator = NULL;
@@ -153,12 +153,12 @@ Error AudioDriverWASAPI::init_render_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;
@@ -184,7 +184,7 @@ Error AudioDriverWASAPI::init_render_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);
 
@@ -205,9 +205,10 @@ Error AudioDriverWASAPI::init_render_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.
@@ -225,7 +226,7 @@ Error AudioDriverWASAPI::init_render_device(bool reinit) {
 		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) {
@@ -236,282 +237,141 @@ Error AudioDriverWASAPI::init_render_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;
-
-	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;
+	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;
 
-		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, 0, p_capture ? REFTIMES_PER_SEC : 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);
 
 	if (OS::get_singleton()->is_stdout_verbose()) {
 		print_line("WASAPI: detected " + itos(channels) + " channels");
 		print_line("WASAPI: audio buffer frames: " + itos(buffer_frames) + " calculated latency: " + itos(buffer_frames * 1000 / mix_rate) + "ms");
 	}
 
-	// Free memory
-	CoTaskMemFree(pwfex);
-
 	return OK;
 }
 
-StringName AudioDriverWASAPI::get_default_capture_device_name(IMMDeviceEnumerator *p_enumerator) {
-	// Setup default device
-	IMMDevice *default_device = NULL;
-	LPWSTR pwszID = NULL;
-	IPropertyStore *props = NULL;
-
-	HRESULT hr = p_enumerator->GetDefaultAudioEndpoint(
-			eCapture, eConsole, &default_device);
-	ERR_FAIL_COND_V(hr != S_OK, "");
-
-	// Get the device ID
-	hr = default_device->GetId(&pwszID);
-	ERR_FAIL_COND_V(hr != S_OK, "");
-
-	// Get the device properties
-	hr = default_device->OpenPropertyStore(
-			STGM_READ, &props);
-	ERR_FAIL_COND_V(hr != S_OK, "");
-
-	PROPVARIANT var_name;
-	PropVariantInit(&var_name);
-
-	// Get the name of the device
-	hr = props->GetValue(PKEY_Device_FriendlyName, &var_name);
-	ERR_FAIL_COND_V(hr != S_OK, "");
-
-	// Return the name of device
-	return String(var_name.pwszVal);
-}
-
-Error AudioDriverWASAPI::init_capture_devices(bool reinit) {
-
-	WAVEFORMATEX *pwfex;
-	IMMDeviceEnumerator *enumerator = NULL;
-	IMMDeviceCollection *device_collection = NULL;
-	IPropertyStore *props = NULL;
-
-	capture_device_id_map.clear();
-
-	HRESULT hr = CoCreateInstance(CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, IID_IMMDeviceEnumerator, (void **)&enumerator);
-	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
+Error AudioDriverWASAPI::init_capture_device(bool reinit) {
 
-	capture_device_default_name = get_default_capture_device_name(enumerator);
+	Error err = audio_device_init(&audio_input, true, reinit);
+	if (err != OK)
+		return err;
 
-	// Enumerate a collection of valid devices
-	hr = enumerator->EnumAudioEndpoints(eCapture, DEVICE_STATE_ACTIVE, &device_collection);
-	ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-	SAFE_RELEASE(enumerator);
-
-	UINT count;
-	hr = device_collection->GetCount(&count);
+	// 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);
 
-	// Loop through the device count
-	for (unsigned int i = 0; i < count; i++) {
-		IMMDevice *device = NULL;
-		LPWSTR pwszID = NULL;
-
-		// Get the device
-		hr = device_collection->Item(i, &device);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		// Get the device ID
-		hr = device->GetId(&pwszID);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		// Get the device properties
-		hr = device->OpenPropertyStore(STGM_READ, &props);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		PROPVARIANT var_name;
-		PropVariantInit(&var_name);
-
-		// Get the name of the device
-		hr = props->GetValue(PKEY_Device_FriendlyName, &var_name);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		// Save the name of device
-		StringName name = String(var_name.pwszVal);
-
-		// DEBUG: print the device name and ID
-		printf("Endpoint %d: \"%S\" (%S)\n", i, var_name.pwszVal, pwszID);
-
-		capture_device_id_map[StringName(pwszID)] = name;
-
-		// Cleanup the ID and properties
-		CoTaskMemFree(pwszID);
-		pwszID = NULL;
-		PropVariantClear(&var_name);
-		SAFE_RELEASE(props)
-
-		// Create a new audio in block descriptor
-		MicrophoneDeviceOutputDirectWASAPI *microphone_device_output_wasapi = memnew(MicrophoneDeviceOutputDirectWASAPI);
-		microphone_device_output_wasapi->name = name;
-		microphone_device_output_wasapi->active = false;
-
-		// Push it into the list and assign it to the hash map for quick access
-		microphone_device_outputs.push_back(microphone_device_output_wasapi);
-		microphone_device_output_map[name] = microphone_device_output_wasapi;
-
-		// Activate the device
-		hr = device->Activate(IID_IAudioClient, CLSCTX_ALL, NULL, (void **)&microphone_device_output_wasapi->audio_client);
-		SAFE_RELEASE(device)
-
-		// Get the sample rate (hz)
-		hr = microphone_device_output_wasapi->audio_client->GetMixFormat(&pwfex);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		microphone_device_output_wasapi->channels = pwfex->nChannels;
-		microphone_device_output_wasapi->mix_rate = pwfex->nSamplesPerSec;
-		microphone_device_output_wasapi->bits_per_sample = pwfex->wBitsPerSample;
-		microphone_device_output_wasapi->frame_size = (microphone_device_output_wasapi->bits_per_sample / 8) * microphone_device_output_wasapi->channels;
-
-		microphone_device_output_wasapi->current_capture_index = 0;
-		microphone_device_output_wasapi->current_capture_size = 0;
-
-		WORD format_tag = pwfex->wFormatTag;
-		if (format_tag == WAVE_FORMAT_EXTENSIBLE) {
-			WAVEFORMATEXTENSIBLE *wfex = (WAVEFORMATEXTENSIBLE *)pwfex;
-
-			if (wfex->SubFormat == KSDATAFORMAT_SUBTYPE_PCM) {
-				format_tag = WAVE_FORMAT_PCM;
-			} else if (wfex->SubFormat == KSDATAFORMAT_SUBTYPE_IEEE_FLOAT) {
-				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) {
-				ERR_PRINT("WASAPI: Format not supported");
-				ERR_FAIL_V(ERR_CANT_OPEN);
-			}
-		}
-		microphone_device_output_wasapi->capture_format_tag = format_tag;
-
-		hr = microphone_device_output_wasapi->audio_client->Initialize(AUDCLNT_SHAREMODE_SHARED, 0, REFTIMES_PER_SEC, 0, pwfex, NULL);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		// Get the max frames
-		UINT32 max_frames;
-		hr = microphone_device_output_wasapi->audio_client->GetBufferSize(&max_frames);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		// Set the buffer size
-		microphone_device_output_wasapi->buffer.resize(max_frames);
-		memset(microphone_device_output_wasapi->buffer.ptrw(), 0x00, microphone_device_output_wasapi->buffer.size() * microphone_device_output_wasapi->frame_size);
-
-		// Get the capture client
-		hr = microphone_device_output_wasapi->audio_client->GetService(IID_IAudioCaptureClient, (void **)&microphone_device_output_wasapi->capture_client);
-		ERR_FAIL_COND_V(hr != S_OK, ERR_CANT_OPEN);
-
-		// TODO: set audio write stream to correct format
-		REFERENCE_TIME hns_actual_duration = (double)REFTIMES_PER_SEC * max_frames / pwfex->nSamplesPerSec;
-
-		// Free memory
-		CoTaskMemFree(pwfex);
-		SAFE_RELEASE(device)
+	// Set the buffer size
+	audio_input_buffer.resize(max_frames * 8);
+	for (int i = 0; i < audio_input_buffer.size(); i++) {
+		audio_input_buffer.write[i] = 0;
 	}
-	SAFE_RELEASE(device_collection)
+	audio_input_position = 0;
 
 	return OK;
 }
 
-Error AudioDriverWASAPI::finish_render_device() {
+Error AudioDriverWASAPI::audio_device_finish(AudioDeviceWASAPI *p_device) {
 
-	if (audio_client) {
-		if (active) {
-			audio_client->Stop();
-			active = false;
+	if (p_device->active) {
+		if (p_device->audio_client) {
+			p_device->audio_client->Stop();
 		}
 
-		audio_client->Release();
-		audio_client = NULL;
+		p_device->active = false;
 	}
 
-	SAFE_RELEASE(render_client)
-	SAFE_RELEASE(audio_client)
+	SAFE_RELEASE(p_device->audio_client)
+	SAFE_RELEASE(p_device->render_client)
+	SAFE_RELEASE(p_device->capture_client)
 
 	return OK;
 }
 
-Error AudioDriverWASAPI::finish_capture_devices() {
+Error AudioDriverWASAPI::finish_render_device() {
 
-	microphone_device_output_map.clear();
-	while (microphone_device_outputs.size() > 0) {
-		MicrophoneDeviceOutputDirectWASAPI *microphone_device_output = static_cast<MicrophoneDeviceOutputDirectWASAPI *>(microphone_device_outputs.get(0));
-		SAFE_RELEASE(microphone_device_output->capture_client)
-		SAFE_RELEASE(microphone_device_output->audio_client)
-		microphone_device_outputs.erase(microphone_device_output);
-		memdelete(microphone_device_output);
-	}
+	return audio_device_finish(&audio_output);
+}
 
-	return OK;
+Error AudioDriverWASAPI::finish_capture_device() {
+
+	return audio_device_finish(&audio_input);
 }
 
 Error AudioDriverWASAPI::init() {
@@ -523,12 +383,6 @@ Error AudioDriverWASAPI::init() {
 		ERR_PRINT("WASAPI: init_render_device error");
 	}
 
-	err = init_capture_devices();
-	if (err != OK) {
-		ERR_PRINT("WASAPI: init_capture_device error");
-	}
-
-	active = false;
 	exit_thread = false;
 	thread_exited = false;
 
@@ -548,7 +402,7 @@ 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;
@@ -561,7 +415,7 @@ 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;
@@ -596,19 +450,28 @@ 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();
 }
 
-float AudioDriverWASAPI::read_sample(WORD format_tag, int bits_per_sample, BYTE *buffer, int i) {
+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) {
@@ -631,19 +494,19 @@ float AudioDriverWASAPI::read_sample(WORD format_tag, int bits_per_sample, BYTE
 				break;
 		}
 
-		return (sample >> 16) / 32768.f;
+		return sample;
 	} else if (format_tag == WAVE_FORMAT_IEEE_FLOAT) {
-		return ((float *)buffer)[i];
+		return int32_t(((float *)buffer)[i] * 32768.0) << 16;
 	} else {
 		ERR_PRINT("WASAPI: Unknown format tag");
 	}
 
-	return 0.f;
+	return 0;
 }
 
-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) {
+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;
@@ -662,157 +525,99 @@ 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) {
-		// Capture
-
-		if (default_capture_device_changed) {
-			if (ad->capture_device_id_map.has(capture_device_id)) {
-				Map<StringName, StringName>::Element *e = ad->capture_device_id_map.find(capture_device_id);
-				ad->lock();
-				ad->start_counting_ticks();
-
-				ad->capture_device_default_name = e->get();
-				ad->update_microphone_default(ad->capture_device_default_name);
-
-				default_capture_device_changed = false;
-
-				ad->stop_counting_ticks();
-				ad->unlock();
-			}
-		}
-
-		for (int i = 0; i < ad->microphone_device_outputs.size(); i++) {
-			MicrophoneDeviceOutputDirectWASAPI *microphone_device_output_wasapi = static_cast<MicrophoneDeviceOutputDirectWASAPI *>(ad->microphone_device_outputs[i]);
-
-			if (microphone_device_output_wasapi->active == false) {
-				continue;
-			}
-
-			UINT32 packet_length = 0;
-			BYTE *data;
-			UINT32 num_frames_available;
-			DWORD flags;
 
-			HRESULT hr = microphone_device_output_wasapi->capture_client->GetNextPacketSize(&packet_length);
-			ERR_BREAK(hr != S_OK);
+		uint32_t read_frames = 0;
+		uint32_t written_frames = 0;
 
-			while (packet_length != 0) {
-				hr = microphone_device_output_wasapi->capture_client->GetBuffer(&data, &num_frames_available, &flags, NULL, NULL);
-				ERR_BREAK(hr != S_OK);
-
-				unsigned int frames_to_copy = num_frames_available;
-
-				if (flags & AUDCLNT_BUFFERFLAGS_SILENT) {
-					memset((char *)(microphone_device_output_wasapi->buffer.ptrw()) + (microphone_device_output_wasapi->current_capture_index * microphone_device_output_wasapi->frame_size), 0, frames_to_copy * microphone_device_output_wasapi->frame_size);
-				} else {
-					// fixme: Only works for floating point atm
-					for (int j = 0; j < frames_to_copy; j++) {
-						float l, r;
-
-						if (microphone_device_output_wasapi->channels == 2) {
-							l = read_sample(microphone_device_output_wasapi->capture_format_tag, microphone_device_output_wasapi->bits_per_sample, data, j * 2);
-							r = read_sample(microphone_device_output_wasapi->capture_format_tag, microphone_device_output_wasapi->bits_per_sample, data, j * 2 + 1);
-						} else if (microphone_device_output_wasapi->channels == 1) {
-							l = r = read_sample(microphone_device_output_wasapi->capture_format_tag, microphone_device_output_wasapi->bits_per_sample, data, j);
-						} else {
-							l = r = 0.f;
-							ERR_PRINT("WASAPI: unsupported channel count in microphone!");
-						}
-
-						microphone_device_output_wasapi->buffer[microphone_device_output_wasapi->current_capture_index++] = AudioFrame(l, r);
+		if (avail_frames == 0) {
+			ad->lock();
+			ad->start_counting_ticks();
 
-						if (microphone_device_output_wasapi->current_capture_index >= microphone_device_output_wasapi->buffer.size()) {
-							microphone_device_output_wasapi->current_capture_index = 0;
-						}
-						if (microphone_device_output_wasapi->current_capture_size < microphone_device_output_wasapi->buffer.size()) {
-							microphone_device_output_wasapi->current_capture_size++;
-						}
-					}
+			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;
 				}
+			}
 
-				hr = microphone_device_output_wasapi->capture_client->ReleaseBuffer(num_frames_available);
-				ERR_BREAK(hr != S_OK);
+			avail_frames = ad->buffer_frames;
+			write_ofs = 0;
 
-				hr = microphone_device_output_wasapi->capture_client->GetNextPacketSize(&packet_length);
-				ERR_BREAK(hr != S_OK);
-			}
+			ad->stop_counting_ticks();
+			ad->unlock();
 		}
 
 		ad->lock();
 		ad->start_counting_ticks();
 
-		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;
-			}
-		}
-
-		ad->stop_counting_ticks();
-		ad->unlock();
-
-		unsigned int left_frames = ad->buffer_frames;
-		unsigned int buffer_idx = 0;
-		while (left_frames > 0 && ad->audio_client) {
-			WaitForSingleObject(ad->event, 1000);
-
-			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");
+						}
+
+						avail_frames -= write_frames;
+						written_frames += write_frames;
+					} else if (hr == AUDCLNT_E_DEVICE_INVALIDATED) {
+						// Device is not valid anymore, reopen it
 
-					left_frames -= write_frames;
-				} else if (hr == AUDCLNT_E_DEVICE_INVALIDATED) {
-					invalidated = true;
-				} else {
-					ERR_PRINT("WASAPI: Get buffer error");
-					ad->exit_thread = true;
+						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;
@@ -829,16 +634,10 @@ void AudioDriverWASAPI::thread_func(void *p_udata) {
 					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_render_device_changed) {
+		if (ad->audio_output.device_name == "Default" && default_render_device_changed) {
 			Error err = ad->finish_render_device();
 			if (err != OK) {
 				ERR_PRINT("WASAPI: finish_render_device error");
@@ -848,23 +647,67 @@ void AudioDriverWASAPI::thread_func(void *p_udata) {
 		}
 
 		// 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;
+		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_render_device error");
 			}
 		}
 
-		if (!ad->audio_client) {
+		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 * ad->audio_input.channels; j++) {
+						int32_t sample;
+
+						if (flags & AUDCLNT_BUFFERFLAGS_SILENT) {
+							sample = 0;
+						} else {
+							sample = read_sample(ad->audio_input.format_tag, ad->audio_input.bits_per_sample, data, j);
+						}
+
+						ad->audio_input_buffer.write[ad->audio_input_position++] = sample;
+						if (ad->audio_input_position >= ad->audio_input_buffer.size()) {
+							ad->audio_input_position = 0;
+						}
+					}
+
+					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);
+				}
+			}
+		}
+
 		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;
@@ -872,12 +715,12 @@ 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;
 		}
 	}
 }
@@ -904,7 +747,7 @@ void AudioDriverWASAPI::finish() {
 		thread = NULL;
 	}
 
-	finish_capture_devices();
+	finish_capture_device();
 	finish_render_device();
 
 	if (mutex) {
@@ -913,83 +756,70 @@ void AudioDriverWASAPI::finish() {
 	}
 }
 
-bool AudioDriverWASAPI::capture_device_start(StringName p_name) {
+Error AudioDriverWASAPI::capture_start() {
 
-	if (microphone_device_output_map.has(p_name)) {
-		MicrophoneDeviceOutputDirectWASAPI *microphone_device_output_wasapi = static_cast<MicrophoneDeviceOutputDirectWASAPI *>(microphone_device_output_map[p_name]);
-		if (microphone_device_output_wasapi->active == false) {
-			microphone_device_output_wasapi->audio_client->Start();
-			microphone_device_output_wasapi->active = true;
-			microphone_device_output_wasapi->set_read_index(-2048);
-		}
+	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 true;
+		return OK;
 	}
 
-	return false;
+	return FAILED;
 }
 
-bool AudioDriverWASAPI::capture_device_stop(StringName p_name) {
+Error AudioDriverWASAPI::capture_stop() {
 
-	if (microphone_device_output_map.has(p_name)) {
-		MicrophoneDeviceOutputDirectWASAPI *microphone_device_output_wasapi = static_cast<MicrophoneDeviceOutputDirectWASAPI *>(microphone_device_output_map[p_name]);
-		if (microphone_device_output_wasapi->active == true) {
-			microphone_device_output_wasapi->audio_client->Stop();
-			microphone_device_output_wasapi->active = false;
-		}
+	if (audio_input.active == true) {
+		audio_input.audio_client->Stop();
+		audio_input.active = false;
 
-		return true;
+		return OK;
 	}
 
-	return false;
+	return FAILED;
 }
 
-PoolStringArray AudioDriverWASAPI::capture_device_get_names() {
+void AudioDriverWASAPI::capture_set_device(StringName p_name) {
 
-	PoolStringArray names;
+	lock();
+	audio_input.new_device = p_name;
+	unlock();
+}
 
-	for (int i = 0; i < microphone_device_outputs.size(); i++) {
-		MicrophoneDeviceOutputDirectWASAPI *microphone_device_output_wasapi = static_cast<MicrophoneDeviceOutputDirectWASAPI *>(microphone_device_outputs.get(i));
-		names.push_back(microphone_device_output_wasapi->name);
-	}
+Array AudioDriverWASAPI::capture_get_device_list() {
 
-	return names;
+	return audio_device_get_list(true);
 }
 
-StringName AudioDriverWASAPI::capture_device_get_default_name() {
+StringName AudioDriverWASAPI::capture_get_device() {
 
 	lock();
-	StringName capture_device_default_name_local = capture_device_default_name;
+	StringName name = audio_input.device_name;
 	unlock();
 
-	return capture_device_default_name_local;
+	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";
-	capture_device_default_name = "";
 }
 
 #endif
diff --git a/drivers/wasapi/audio_driver_wasapi.h b/drivers/wasapi/audio_driver_wasapi.h
index e722d85353..a131eb5dae 100644
--- a/drivers/wasapi/audio_driver_wasapi.h
+++ b/drivers/wasapi/audio_driver_wasapi.h
@@ -43,53 +43,63 @@
 
 class AudioDriverWASAPI : public AudioDriver {
 
-	HANDLE event;
-	// Audio out
-	IAudioClient *audio_client;
-	IAudioRenderClient *render_client;
-	// Microphone
-	class MicrophoneDeviceOutputDirectWASAPI : public MicrophoneDeviceOutputDirect {
+	class AudioDeviceWASAPI {
 	public:
 		IAudioClient *audio_client;
+		IAudioRenderClient *render_client;
 		IAudioCaptureClient *capture_client;
-		WORD capture_format_tag;
+		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";
+		}
 	};
-	//
-	Mutex *mutex;
-	Thread *thread;
 
-	String device_name;
-	String new_device;
-	String capture_device_default_name;
+	AudioDeviceWASAPI audio_input;
+	AudioDeviceWASAPI audio_output;
 
-	WORD format_tag;
-	WORD bits_per_sample;
+	Mutex *mutex;
+	Thread *thread;
 
 	Vector<int32_t> samples_in;
 
-	Map<StringName, StringName> capture_device_id_map;
-
-	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_ float read_sample(WORD format_tag, int bits_per_sample, BYTE *buffer, int i);
+	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);
 
-	StringName get_default_capture_device_name(IMMDeviceEnumerator *p_enumerator);
-
 	Error init_render_device(bool reinit = false);
-	Error init_capture_devices(bool reinit = false);
+	Error init_capture_device(bool reinit = false);
 
 	Error finish_render_device();
-	Error finish_capture_devices();
+	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 {
@@ -107,10 +117,11 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
+	virtual Error capture_start();
+	virtual Error capture_stop();
+	virtual Array capture_get_device_list();
+	virtual void capture_set_device(StringName p_name);
+	virtual StringName capture_get_device();
 
 	AudioDriverWASAPI();
 };
diff --git a/drivers/xaudio2/audio_driver_xaudio2.cpp b/drivers/xaudio2/audio_driver_xaudio2.cpp
index 2a320b620e..a1002ef4f9 100644
--- a/drivers/xaudio2/audio_driver_xaudio2.cpp
+++ b/drivers/xaudio2/audio_driver_xaudio2.cpp
@@ -210,28 +210,6 @@ void AudioDriverXAudio2::finish() {
 	thread = NULL;
 };
 
-bool AudioDriverXAudio2::capture_device_start(StringName p_name) {
-
-	return false;
-}
-
-bool AudioDriverXAudio2::capture_device_stop(StringName p_name) {
-
-	return false;
-}
-
-PoolStringArray AudioDriverXAudio2::capture_device_get_names() {
-
-	PoolStringArray names;
-
-	return names;
-}
-
-StringName AudioDriverXAudio2::capture_device_get_default_name() {
-
-	return "";
-}
-
 AudioDriverXAudio2::AudioDriverXAudio2() {
 
 	mutex = NULL;
diff --git a/drivers/xaudio2/audio_driver_xaudio2.h b/drivers/xaudio2/audio_driver_xaudio2.h
index b78ab136da..42e1adb2b7 100644
--- a/drivers/xaudio2/audio_driver_xaudio2.h
+++ b/drivers/xaudio2/audio_driver_xaudio2.h
@@ -103,11 +103,6 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	AudioDriverXAudio2();
 	~AudioDriverXAudio2();
 };
diff --git a/platform/android/audio_driver_jandroid.cpp b/platform/android/audio_driver_jandroid.cpp
index 561ad106e4..b9f1f1eab0 100644
--- a/platform/android/audio_driver_jandroid.cpp
+++ b/platform/android/audio_driver_jandroid.cpp
@@ -195,28 +195,6 @@ void AudioDriverAndroid::finish() {
 	active = false;
 }
 
-bool AudioDriverAndroid::capture_device_start(StringName p_name) {
-
-	return false;
-}
-
-bool AudioDriverAndroid::capture_device_stop(StringName p_name) {
-
-	return false;
-}
-
-PoolStringArray AudioDriverAndroid::capture_device_get_names() {
-
-	PoolStringArray names;
-
-	return names;
-}
-
-StringName AudioDriverAndroid::capture_device_get_default_name() {
-
-	return "";
-}
-
 void AudioDriverAndroid::set_pause(bool p_pause) {
 
 	JNIEnv *env = ThreadAndroid::get_env();
diff --git a/platform/android/audio_driver_jandroid.h b/platform/android/audio_driver_jandroid.h
index a54f0cae6a..763f0e9b5a 100644
--- a/platform/android/audio_driver_jandroid.h
+++ b/platform/android/audio_driver_jandroid.h
@@ -70,11 +70,6 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	virtual void set_pause(bool p_pause);
 
 	static void setup(jobject p_io);
diff --git a/platform/android/audio_driver_opensl.cpp b/platform/android/audio_driver_opensl.cpp
index 38f19d41e9..28e3ea962f 100644
--- a/platform/android/audio_driver_opensl.cpp
+++ b/platform/android/audio_driver_opensl.cpp
@@ -249,28 +249,6 @@ void AudioDriverOpenSL::finish() {
 	(*sl)->Destroy(sl);
 }
 
-bool AudioDriverOpenSL::capture_device_start(StringName p_name) {
-
-	return false;
-}
-
-bool AudioDriverOpenSL::capture_device_stop(StringName p_name) {
-
-	return false;
-}
-
-PoolStringArray AudioDriverOpenSL::capture_device_get_names() {
-
-	PoolStringArray names;
-
-	return names;
-}
-
-StringName AudioDriverOpenSL::capture_device_get_default_name() {
-
-	return "";
-}
-
 void AudioDriverOpenSL::set_pause(bool p_pause) {
 
 	pause = p_pause;
diff --git a/platform/android/audio_driver_opensl.h b/platform/android/audio_driver_opensl.h
index f3f1939733..88cb122414 100644
--- a/platform/android/audio_driver_opensl.h
+++ b/platform/android/audio_driver_opensl.h
@@ -98,11 +98,6 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	virtual void set_pause(bool p_pause);
 
 	AudioDriverOpenSL();
diff --git a/servers/audio/audio_driver_dummy.cpp b/servers/audio/audio_driver_dummy.cpp
index 7074dbca3a..be36c3b748 100644
--- a/servers/audio/audio_driver_dummy.cpp
+++ b/servers/audio/audio_driver_dummy.cpp
@@ -125,28 +125,6 @@ void AudioDriverDummy::finish() {
 	thread = NULL;
 };
 
-bool AudioDriverDummy::capture_device_start(StringName p_name) {
-
-	return false;
-}
-
-bool AudioDriverDummy::capture_device_stop(StringName p_name) {
-
-	return false;
-}
-
-PoolStringArray AudioDriverDummy::capture_device_get_names() {
-
-	PoolStringArray names;
-
-	return names;
-}
-
-StringName AudioDriverDummy::capture_device_get_default_name() {
-
-	return "";
-}
-
 AudioDriverDummy::AudioDriverDummy() {
 
 	mutex = NULL;
diff --git a/servers/audio/audio_driver_dummy.h b/servers/audio/audio_driver_dummy.h
index 48e8c731ea..ba4f69edf4 100644
--- a/servers/audio/audio_driver_dummy.h
+++ b/servers/audio/audio_driver_dummy.h
@@ -68,11 +68,6 @@ public:
 	virtual void unlock();
 	virtual void finish();
 
-	virtual bool capture_device_start(StringName p_name);
-	virtual bool capture_device_stop(StringName p_name);
-	virtual PoolStringArray capture_device_get_names();
-	virtual StringName capture_device_get_default_name();
-
 	AudioDriverDummy();
 	~AudioDriverDummy();
 };
diff --git a/servers/audio/audio_stream.cpp b/servers/audio/audio_stream.cpp
index 206f1861a3..0c8a1810d2 100644
--- a/servers/audio/audio_stream.cpp
+++ b/servers/audio/audio_stream.cpp
@@ -29,6 +29,7 @@
 /*************************************************************************/
 
 #include "audio_stream.h"
+#include "os/os.h"
 
 //////////////////////////////
 
@@ -119,33 +120,11 @@ String AudioStreamMicrophone::get_stream_name() const {
 	return "Microphone";
 }
 
-void AudioStreamMicrophone::set_microphone_name(const String &p_name) {
-	if (microphone_name != p_name) {
-		microphone_name = p_name;
-
-		for (Set<AudioStreamPlaybackMicrophone *>::Element *E = playbacks.front(); E; E = E->next()) {
-			if (E->get()->active) {
-				// Is this the right thing to do?
-				E->get()->stop();
-				E->get()->start();
-			}
-		}
-	}
-}
-
-StringName AudioStreamMicrophone::get_microphone_name() const {
-	return microphone_name;
-}
-
 float AudioStreamMicrophone::get_length() const {
 	return 0;
 }
 
 void AudioStreamMicrophone::_bind_methods() {
-	ClassDB::bind_method(D_METHOD("set_microphone_name", "name"), &AudioStreamMicrophone::set_microphone_name);
-	ClassDB::bind_method(D_METHOD("get_microphone_name"), &AudioStreamMicrophone::get_microphone_name);
-
-	ADD_PROPERTY(PropertyInfo(Variant::STRING, "microphone_name"), "set_microphone_name", "get_microphone_name");
 }
 
 AudioStreamMicrophone::AudioStreamMicrophone() {
@@ -153,20 +132,25 @@ AudioStreamMicrophone::AudioStreamMicrophone() {
 
 void AudioStreamPlaybackMicrophone::_mix_internal(AudioFrame *p_buffer, int p_frames) {
 
-	AudioDriver::MicrophoneDeviceOutput *microphone_device_output = reciever->owner;
-	const Vector<AudioFrame> &source_buffer = microphone_device_output->get_buffer();
-	int current_buffer_size = microphone_device_output->get_current_buffer_size();
+	AudioDriver::get_singleton()->lock();
+
+	Vector<int32_t> buf = AudioDriver::get_singleton()->get_audio_input_buffer();
 
 	for (int i = 0; i < p_frames; i++) {
-		if (current_buffer_size >= internal_mic_offset) {
-			if (internal_mic_offset >= source_buffer.size()) {
-				internal_mic_offset = 0;
-			}
-			p_buffer[i] = source_buffer[internal_mic_offset++];
-		} else {
-			p_buffer[i] = AudioFrame(0.f, 0.f);
+
+		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);
 	}
+
+	AudioDriver::get_singleton()->unlock();
 }
 
 void AudioStreamPlaybackMicrophone::mix(AudioFrame *p_buffer, float p_rate_scale, int p_frames) {
@@ -174,28 +158,21 @@ void AudioStreamPlaybackMicrophone::mix(AudioFrame *p_buffer, float p_rate_scale
 }
 
 float AudioStreamPlaybackMicrophone::get_stream_sampling_rate() {
-	return reciever->owner->get_mix_rate();
+	return AudioDriver::get_singleton()->get_mix_rate();
 }
 
 void AudioStreamPlaybackMicrophone::start(float p_from_pos) {
-	internal_mic_offset = 0;
-	active = true;
+	input_ofs = 0;
 
-	// note: can this be called twice?
-	reciever = AudioServer::get_singleton()->create_microphone_reciever(microphone->get_microphone_name());
-	if (reciever == NULL) {
-		active = false;
-	}
+	AudioDriver::get_singleton()->capture_start();
 
+	active = true;
 	_begin_resample();
 }
 
 void AudioStreamPlaybackMicrophone::stop() {
+	AudioDriver::get_singleton()->capture_stop();
 	active = false;
-	if (reciever != NULL) {
-		AudioServer::get_singleton()->destroy_microphone_reciever(reciever);
-		reciever = NULL;
-	}
 }
 
 bool AudioStreamPlaybackMicrophone::is_playing() const {
@@ -220,8 +197,6 @@ AudioStreamPlaybackMicrophone::~AudioStreamPlaybackMicrophone() {
 }
 
 AudioStreamPlaybackMicrophone::AudioStreamPlaybackMicrophone() {
-	internal_mic_offset = 0;
-	reciever = NULL;
 }
 
 ////////////////////////////////
diff --git a/servers/audio/audio_stream.h b/servers/audio/audio_stream.h
index 352cb8c507..f0f756129e 100644
--- a/servers/audio/audio_stream.h
+++ b/servers/audio/audio_stream.h
@@ -104,7 +104,6 @@ class AudioStreamMicrophone : public AudioStream {
 	friend class AudioStreamPlaybackMicrophone;
 
 	Set<AudioStreamPlaybackMicrophone *> playbacks;
-	StringName microphone_name;
 
 protected:
 	static void _bind_methods();
@@ -113,9 +112,6 @@ public:
 	virtual Ref<AudioStreamPlayback> instance_playback();
 	virtual String get_stream_name() const;
 
-	void set_microphone_name(const String &p_name);
-	StringName get_microphone_name() const;
-
 	virtual float get_length() const; //if supported, otherwise return 0
 
 	AudioStreamMicrophone();
@@ -127,10 +123,9 @@ class AudioStreamPlaybackMicrophone : public AudioStreamPlaybackResampled {
 	friend class AudioStreamMicrophone;
 
 	bool active;
-	uint32_t internal_mic_offset;
+	unsigned int input_ofs;
 
 	Ref<AudioStreamMicrophone> microphone;
-	AudioDriver::MicrophoneReciever *reciever;
 
 protected:
 	virtual void _mix_internal(AudioFrame *p_buffer, int p_frames);
diff --git a/servers/audio_server.cpp b/servers/audio_server.cpp
index 10fa69a12e..eb034fa615 100644
--- a/servers/audio_server.cpp
+++ b/servers/audio_server.cpp
@@ -73,42 +73,6 @@ void AudioDriver::update_mix_time(int p_frames) {
 		_last_mix_time = OS::get_singleton()->get_ticks_usec();
 }
 
-void AudioDriver::update_microphone_default(StringName p_device_name) {
-	if (default_microphone_device_output != NULL) {
-		MicrophoneDeviceOutput *output = default_microphone_device_output->owner;
-		output->remove_reciever(default_microphone_device_output);
-
-		while (output != NULL) {
-			MicrophoneDeviceOutput *owner = output->owner;
-			if (output->recievers.size() == 0) {
-				if (owner == NULL) {
-					if (output->active == true) {
-						capture_device_stop(output->name);
-						output->active == false;
-					}
-				} else {
-					owner->remove_reciever(output);
-					memdelete(output);
-				}
-				output = owner;
-			}
-		}
-
-		if (microphone_device_output_map.has(p_device_name)) {
-
-			Map<StringName, MicrophoneDeviceOutput *>::Element *e = microphone_device_output_map.find(p_device_name);
-			MicrophoneDeviceOutput *new_output = e->get();
-			new_output->add_reciever(default_microphone_device_output);
-			if (new_output->active == false) {
-				capture_device_start(p_device_name);
-				new_output->active = true;
-			}
-		}
-
-		output = default_microphone_device_output->owner;
-	}
-}
-
 double AudioDriver::get_mix_time() const {
 
 	double total = (OS::get_singleton()->get_ticks_usec() - _last_mix_time) / 1000000.0;
@@ -138,75 +102,19 @@ int AudioDriver::get_total_channels_by_speaker_mode(AudioDriver::SpeakerMode p_m
 	ERR_FAIL_V(2);
 }
 
-AudioDriver::MicrophoneReciever *AudioDriver::create_microphone_reciever(const StringName &p_device_name) {
-
-	MicrophoneReciever *microphone_reciever = NULL;
-	MicrophoneDeviceOutput *reciever_output = NULL;
-	MicrophoneDeviceOutput *device_output = NULL;
-
-	StringName device_name = capture_device_get_default_name();
-
-	if (microphone_device_output_map.has(device_name)) {
-
-		Map<StringName, MicrophoneDeviceOutput *>::Element *e = microphone_device_output_map.find(device_name);
-		device_output = e->get();
-	}
-
-	if (device_output) {
-		if (p_device_name == "") {
-			if (default_microphone_device_output != NULL) {
-				reciever_output = default_microphone_device_output;
-			} else {
-				// Default reciever does not exist, create it and connect it
-				default_microphone_device_output = memnew(MicrophoneDeviceOutputIndirect);
-				reciever_output = default_microphone_device_output;
-				device_output->add_reciever(reciever_output);
-			}
-		} else {
-			if (microphone_device_output_map.has(p_device_name)) {
-				reciever_output = device_output;
-			}
-		}
+Array AudioDriver::get_device_list() {
+	Array list;
 
-		if (reciever_output) {
-			microphone_reciever = memnew(MicrophoneReciever);
-			reciever_output->add_reciever(microphone_reciever);
-			if (device_output->active == false) {
-				capture_device_start(device_name);
-				device_output->active = true;
-			}
-		}
-	}
+	list.push_back("Default");
 
-	return microphone_reciever;
+	return list;
 }
 
-void AudioDriver::destroy_microphone_reciever(AudioDriver::MicrophoneReciever *p_microphone_reciever) {
-
-	if (p_microphone_reciever != NULL) {
-		MicrophoneDeviceOutput *output = p_microphone_reciever->owner;
-		output->remove_reciever(p_microphone_reciever);
-
-		while (output != NULL) {
-			MicrophoneDeviceOutput *owner = output->owner;
-			if (output->recievers.size() == 0) {
-				if (owner == NULL) {
-					if (output->active == true) {
-						capture_device_stop(output->name);
-						output->active == false;
-					}
-				} else {
-					owner->remove_reciever(output);
-					memdelete(output);
-				}
-				output = owner;
-			}
-		}
-		memdelete(p_microphone_reciever);
-	}
+String AudioDriver::get_device() {
+	return "Default";
 }
 
-Array AudioDriver::get_device_list() {
+Array AudioDriver::capture_get_device_list() {
 	Array list;
 
 	list.push_back("Default");
@@ -214,17 +122,11 @@ Array AudioDriver::get_device_list() {
 	return list;
 }
 
-String AudioDriver::get_device() {
-	return "Default";
-}
-
 AudioDriver::AudioDriver() {
 
 	_last_mix_time = 0;
 	_mix_amount = 0;
 
-	default_microphone_device_output = NULL;
-
 #ifdef DEBUG_ENABLED
 	prof_time = 0;
 #endif
@@ -1308,32 +1210,19 @@ void AudioServer::set_device(String device) {
 	AudioDriver::get_singleton()->set_device(device);
 }
 
-PoolStringArray AudioServer::audio_in_get_device_names() {
-
-	lock();
-	PoolStringArray device_names = AudioDriver::get_singleton()->capture_device_get_names();
-	unlock();
+Array AudioServer::capture_get_device_list() {
 
-	return device_names;
+	return AudioDriver::get_singleton()->capture_get_device_list();
 }
 
-AudioDriver::MicrophoneReciever *AudioServer::create_microphone_reciever(const StringName &p_device_name) {
-	AudioDriver::MicrophoneReciever *microphone_reciever = NULL;
-
-	lock();
-	microphone_reciever = AudioDriver::get_singleton()->create_microphone_reciever(p_device_name);
-	unlock();
+StringName AudioServer::capture_get_device() {
 
-	return microphone_reciever;
+	return AudioDriver::get_singleton()->capture_get_device();
 }
 
-void AudioServer::destroy_microphone_reciever(AudioDriver::MicrophoneReciever *p_microphone_reciever) {
-	lock();
-	AudioDriver::get_singleton()->destroy_microphone_reciever(p_microphone_reciever);
-	unlock();
-}
+void AudioServer::capture_set_device(StringName device) {
 
-void AudioServer::_change_default_device(StringName p_recording_device_default_name) {
+	AudioDriver::get_singleton()->capture_set_device(device);
 }
 
 void AudioServer::_bind_methods() {
diff --git a/servers/audio_server.h b/servers/audio_server.h
index 68a56c38f3..035942cf89 100644
--- a/servers/audio_server.h
+++ b/servers/audio_server.h
@@ -53,9 +53,11 @@ class AudioDriver {
 #endif
 
 protected:
+	Vector<int32_t> audio_input_buffer;
+	unsigned int audio_input_position;
+
 	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 update_microphone_default(StringName p_device_name);
 
 #ifdef DEBUG_ENABLED
 	_FORCE_INLINE_ void start_counting_ticks() { prof_ticks = OS::get_singleton()->get_ticks_usec(); }
@@ -68,126 +70,6 @@ protected:
 public:
 	double get_mix_time() const; //useful for video -> audio sync
 
-	class MicrophoneDeviceOutput;
-
-	class MicrophoneReciever {
-	public:
-		MicrophoneDeviceOutput *owner;
-
-		MicrophoneReciever() {
-			owner = NULL;
-		}
-
-		~MicrophoneReciever() {
-		}
-	};
-
-	class MicrophoneDeviceOutput : public MicrophoneReciever {
-	public:
-		StringName name;
-		bool active;
-		Vector<MicrophoneReciever *> recievers;
-
-		virtual unsigned int get_mix_rate() = 0;
-		virtual Vector<AudioFrame> &get_buffer() = 0;
-		virtual int get_current_buffer_size() = 0;
-		virtual int get_read_index() = 0;
-		virtual void set_read_index(int p_temp_index) = 0;
-
-		void add_reciever(MicrophoneReciever *p_reciever) {
-			if (p_reciever == NULL) {
-				ERR_PRINT("Attempted to add NULL reciever")
-				return;
-			}
-			if (recievers.find(p_reciever) == -1) {
-				recievers.push_back(p_reciever);
-				p_reciever->owner = this;
-			} else {
-				ERR_PRINT("Duplicate reciever added")
-			}
-		}
-
-		void remove_reciever(MicrophoneReciever *p_reciever) {
-			if (p_reciever == NULL) {
-				ERR_PRINT("Attempted to remove NULL reciever")
-				return;
-			}
-
-			int index = recievers.find(p_reciever);
-
-			if (index != -1) {
-				recievers.remove(index);
-				p_reciever->owner = NULL;
-			} else {
-				ERR_PRINT("Attempted to remove invalid reciever")
-			}
-		}
-	};
-
-	class MicrophoneDeviceOutputDirect : public MicrophoneDeviceOutput {
-	public:
-		enum MicrophoneFormat {
-			FORMAT_FLOAT,
-			FORMAT_PCM
-		};
-
-		MicrophoneFormat microphone_format;
-		unsigned short bits_per_sample;
-		unsigned int channels;
-		unsigned int mix_rate;
-		unsigned short frame_size;
-		int read_index = -2048;
-
-		unsigned int current_capture_index;
-		unsigned int current_capture_size;
-		Vector<AudioFrame> buffer;
-
-		unsigned int get_mix_rate() {
-			return mix_rate;
-		};
-
-		Vector<AudioFrame> &get_buffer() {
-			return buffer;
-		};
-
-		int get_current_buffer_size() { return current_capture_size; }
-
-		int get_read_index() {
-			return read_index;
-		}
-
-		void set_read_index(int p_read_index) {
-			read_index = p_read_index;
-		}
-	};
-
-	class MicrophoneDeviceOutputIndirect : public MicrophoneDeviceOutput {
-	public:
-		unsigned int get_mix_rate() {
-			return owner->get_mix_rate();
-		};
-
-		Vector<AudioFrame> &get_buffer() {
-			return owner->get_buffer();
-		};
-
-		int get_read_index() {
-			return owner->get_read_index();
-		}
-
-		void set_read_index(int p_read_index) {
-			owner->set_read_index(p_read_index);
-		}
-		int get_current_buffer_size() { return owner->get_current_buffer_size(); }
-	};
-
-	MicrophoneDeviceOutputIndirect *default_microphone_device_output;
-
-	Vector<MicrophoneDeviceOutput *> microphone_device_outputs;
-	Map<StringName, MicrophoneDeviceOutput *> microphone_device_output_map;
-
-	Vector<MicrophoneReciever *> direct_recievers;
-
 	enum SpeakerMode {
 		SPEAKER_MODE_STEREO,
 		SPEAKER_SURROUND_31,
@@ -214,18 +96,19 @@ public:
 	virtual void unlock() = 0;
 	virtual void finish() = 0;
 
-	virtual bool capture_device_start(StringName p_name) = 0;
-	virtual bool capture_device_stop(StringName p_name) = 0;
-	virtual PoolStringArray capture_device_get_names() = 0;
-	virtual StringName capture_device_get_default_name() = 0;
+	virtual Error capture_start() { return FAILED; }
+	virtual Error capture_stop() { return FAILED; }
+	virtual void capture_set_device(StringName p_name) {}
+	virtual StringName 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;
 
-	AudioDriver::MicrophoneReciever *create_microphone_reciever(const StringName &p_device_name);
-	void destroy_microphone_reciever(AudioDriver::MicrophoneReciever *p_microphone_reciever);
+	Vector<int32_t> get_audio_input_buffer() { return audio_input_buffer; }
+	unsigned int get_audio_input_position() { return audio_input_position; }
 
 #ifdef DEBUG_ENABLED
 	uint64_t get_profiling_time() const { return prof_time; }
@@ -379,7 +262,6 @@ private:
 
 	friend class AudioDriver;
 	void _driver_process(int p_frames, int32_t *p_buffer);
-	void _change_default_device(StringName p_recording_device_default_name);
 
 protected:
 	static void _bind_methods();
@@ -479,10 +361,9 @@ public:
 	String get_device();
 	void set_device(String device);
 
-	AudioDriver::MicrophoneReciever *create_microphone_reciever(const StringName &p_device_name);
-	void destroy_microphone_reciever(AudioDriver::MicrophoneReciever *p_microphone_reciever);
-
-	PoolStringArray audio_in_get_device_names();
+	Array capture_get_device_list();
+	StringName capture_get_device();
+	void capture_set_device(StringName device);
 
 	float get_output_latency() { return output_latency; }
 	AudioServer();