Merge pull request #19106 from SaracenOne/audio_mic

[WIP] Experimental microphone support

2 files changed, 429 insertions, 166 deletions

diff --git a/drivers/wasapi/audio_driver_wasapi.cpp b/drivers/wasapi/audio_driver_wasapi.cpp
index 5982955c4f..a2f619a6ef 100644
--- a/drivers/wasapi/audio_driver_wasapi.cpp
+++ b/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
diff --git a/drivers/wasapi/audio_driver_wasapi.h b/drivers/wasapi/audio_driver_wasapi.h
index f3ee5976eb..3d94f3ba49 100644
--- a/drivers/wasapi/audio_driver_wasapi.h
+++ b/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();
 };