wasapi: Allow opening devices in exclusive mode

This provides much lower latency compared to opening in shared mode,
but it also means that the device cannot be opened by any other
application. The advantage is that the achievable latency is much
lower.

In shared mode, WASAPI's engine period is 10ms, and so that is the
lowest latency achievable.

In exclusive mode, the limit is the device period itself, which in my
testing with USB DACs, on-board PCI sound-cards, and HDMI cards is
between 2ms and 3.33ms.

We set our audioringbuffer limits to match the device, so the
achievable sink latency is 6-9ms. Further improvements can be made if
needed.

https://bugzilla.gnome.org/show_bug.cgi?id=793289

6 files changed, 232 insertions, 68 deletions

diff --git a/sys/wasapi/gstwasapisink.c b/sys/wasapi/gstwasapisink.c
index a124ba30a..0f0d299d4 100644
--- a/sys/wasapi/gstwasapisink.c
+++ b/sys/wasapi/gstwasapisink.c
@@ -40,8 +40,6 @@
 
 #include "gstwasapisink.h"
 
-#include <mmdeviceapi.h>
-
 GST_DEBUG_CATEGORY_STATIC (gst_wasapi_sink_debug);
 #define GST_CAT_DEFAULT gst_wasapi_sink_debug
 
@@ -50,15 +48,17 @@ static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
     GST_PAD_ALWAYS,
     GST_STATIC_CAPS (GST_WASAPI_STATIC_CAPS));
 
-#define DEFAULT_ROLE    GST_WASAPI_DEVICE_ROLE_CONSOLE
-#define DEFAULT_MUTE    FALSE
+#define DEFAULT_ROLE      GST_WASAPI_DEVICE_ROLE_CONSOLE
+#define DEFAULT_MUTE      FALSE
+#define DEFAULT_EXCLUSIVE FALSE
 
 enum
 {
   PROP_0,
   PROP_ROLE,
   PROP_MUTE,
-  PROP_DEVICE
+  PROP_DEVICE,
+  PROP_EXCLUSIVE
 };
 
 static void gst_wasapi_sink_dispose (GObject * object);
@@ -116,6 +116,12 @@ gst_wasapi_sink_class_init (GstWasapiSinkClass * klass)
           "WASAPI playback device as a GUID string",
           NULL, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
 
+  g_object_class_install_property (gobject_class,
+      PROP_EXCLUSIVE,
+      g_param_spec_boolean ("exclusive", "Exclusive mode",
+          "Open the device in exclusive mode",
+          DEFAULT_EXCLUSIVE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+
   gst_element_class_add_static_pad_template (gstelement_class, &sink_template);
   gst_element_class_set_static_metadata (gstelement_class, "WasapiSrc",
       "Sink/Audio",
@@ -209,6 +215,10 @@ gst_wasapi_sink_set_property (GObject * object, guint prop_id,
           device ? g_utf8_to_utf16 (device, -1, NULL, NULL, NULL) : NULL;
       break;
     }
+    case PROP_EXCLUSIVE:
+      self->sharemode = g_value_get_boolean (value)
+          ? AUDCLNT_SHAREMODE_EXCLUSIVE : AUDCLNT_SHAREMODE_SHARED;
+      break;
     default:
       G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
       break;
@@ -232,6 +242,10 @@ gst_wasapi_sink_get_property (GObject * object, guint prop_id,
       g_value_take_string (value, self->device_strid ?
           g_utf16_to_utf8 (self->device_strid, -1, NULL, NULL, NULL) : NULL);
       break;
+    case PROP_EXCLUSIVE:
+      g_value_set_boolean (value,
+          self->sharemode == AUDCLNT_SHAREMODE_EXCLUSIVE);
+      break;
     default:
       G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
       break;
@@ -244,7 +258,6 @@ gst_wasapi_sink_get_caps (GstBaseSink * bsink, GstCaps * filter)
   GstWasapiSink *self = GST_WASAPI_SINK (bsink);
   WAVEFORMATEX *format = NULL;
   GstCaps *caps = NULL;
-  HRESULT hr;
 
   GST_DEBUG_OBJECT (self, "entering get caps");
 
@@ -252,16 +265,18 @@ gst_wasapi_sink_get_caps (GstBaseSink * bsink, GstCaps * filter)
     caps = gst_caps_ref (self->cached_caps);
   } else {
     GstCaps *template_caps;
+    gboolean ret;
 
     template_caps = gst_pad_get_pad_template_caps (bsink->sinkpad);
 
     if (!self->client)
       gst_wasapi_sink_open (GST_AUDIO_SINK (bsink));
 
-    hr = IAudioClient_GetMixFormat (self->client, &format);
-    if (hr != S_OK || format == NULL) {
+    ret = gst_wasapi_util_get_device_format (GST_ELEMENT (self),
+        self->sharemode, self->device, self->client, &format);
+    if (!ret) {
       GST_ELEMENT_ERROR (self, STREAM, FORMAT, (NULL),
-          ("GetMixFormat failed: %s", gst_wasapi_util_hresult_to_string (hr)));
+          ("failed to detect format"));
       goto out;
     }
 
@@ -302,6 +317,7 @@ gst_wasapi_sink_open (GstAudioSink * asink)
 {
   GstWasapiSink *self = GST_WASAPI_SINK (asink);
   gboolean res = FALSE;
+  IMMDevice *device = NULL;
   IAudioClient *client = NULL;
 
   GST_DEBUG_OBJECT (self, "opening device");
@@ -314,7 +330,7 @@ gst_wasapi_sink_open (GstAudioSink * asink)
    * For example, perhaps we should automatically switch to the new device if
    * the default device is changed and a device isn't explicitly selected. */
   if (!gst_wasapi_util_get_device_client (GST_ELEMENT (self), FALSE,
-          self->role, self->device_strid, &client)) {
+          self->role, self->device_strid, &device, &client)) {
     if (!self->device_strid)
       GST_ELEMENT_ERROR (self, RESOURCE, OPEN_WRITE, (NULL),
           ("Failed to get default device"));
@@ -325,6 +341,7 @@ gst_wasapi_sink_open (GstAudioSink * asink)
   }
 
   self->client = client;
+  self->device = device;
   res = TRUE;
 
 beach:
@@ -337,6 +354,11 @@ gst_wasapi_sink_close (GstAudioSink * asink)
 {
   GstWasapiSink *self = GST_WASAPI_SINK (asink);
 
+  if (self->device != NULL) {
+    IUnknown_Release (self->device);
+    self->device = NULL;
+  }
+
   if (self->client != NULL) {
     IUnknown_Release (self->client);
     self->client = NULL;
@@ -352,7 +374,7 @@ gst_wasapi_sink_prepare (GstAudioSink * asink, GstAudioRingBufferSpec * spec)
   gboolean res = FALSE;
   REFERENCE_TIME latency_rt;
   IAudioRenderClient *render_client = NULL;
-  gint64 default_period, min_period;
+  gint64 default_period, min_period, use_period;
   guint bpf, rate;
   HRESULT hr;
 
@@ -364,15 +386,27 @@ gst_wasapi_sink_prepare (GstAudioSink * asink, GstAudioRingBufferSpec * spec)
   GST_INFO_OBJECT (self, "wasapi default period: %" G_GINT64_FORMAT
       ", min period: %" G_GINT64_FORMAT, default_period, min_period);
 
-  /* Set hnsBufferDuration to 0, which should, in theory, tell the device to
-   * create a buffer with the smallest latency possible. In practice, this is
-   * usually 2 * default_period. See:
-   * https://msdn.microsoft.com/en-us/library/windows/desktop/dd370871(v=vs.85).aspx
-   *
-   * NOTE: min_period is a lie, and I have never seen WASAPI use it as the
-   * current period */
-  hr = IAudioClient_Initialize (self->client, AUDCLNT_SHAREMODE_SHARED,
-      AUDCLNT_STREAMFLAGS_EVENTCALLBACK, 0, 0, self->mix_format, NULL);
+  if (self->sharemode == AUDCLNT_SHAREMODE_SHARED) {
+    use_period = default_period;
+    /* Set hnsBufferDuration to 0, which should, in theory, tell the device to
+     * create a buffer with the smallest latency possible. In practice, this is
+     * usually 2 * default_period. See:
+     * https://msdn.microsoft.com/en-us/library/windows/desktop/dd370871(v=vs.85).aspx
+     *
+     * NOTE: min_period is a lie, and I have never seen WASAPI use it as the
+     * current period */
+    hr = IAudioClient_Initialize (self->client, AUDCLNT_SHAREMODE_SHARED,
+        AUDCLNT_STREAMFLAGS_EVENTCALLBACK, 0, 0, self->mix_format, NULL);
+  } else {
+    use_period = min_period;
+    /* For some reason, we need to call this another time for exclusive mode */
+    CoInitialize (NULL);
+    /* FIXME: We should be able to use min_period as the device buffer size,
+     * but I'm hitting a problem in GStreamer. */
+    hr = IAudioClient_Initialize (self->client, AUDCLNT_SHAREMODE_EXCLUSIVE,
+        AUDCLNT_STREAMFLAGS_EVENTCALLBACK, use_period, use_period,
+        self->mix_format, NULL);
+  }
   if (hr != S_OK) {
     GST_ELEMENT_ERROR (self, RESOURCE, OPEN_READ, (NULL),
         ("IAudioClient::Initialize () failed: %s",
@@ -394,8 +428,10 @@ gst_wasapi_sink_prepare (GstAudioSink * asink, GstAudioRingBufferSpec * spec)
 
   /* Actual latency-time/buffer-time are different now */
   spec->segsize = gst_util_uint64_scale_int_round (rate * bpf,
-      default_period * 100, GST_SECOND);
-  spec->segtotal = (self->buffer_frame_count * bpf) / spec->segsize;
+      use_period * 100, GST_SECOND);
+
+  /* We need a minimum of 2 segments to ensure glitch-free playback */
+  spec->segtotal = MAX (self->buffer_frame_count * bpf / spec->segsize, 2);
 
   GST_INFO_OBJECT (self, "segsize is %i, segtotal is %i", spec->segsize,
       spec->segtotal);
@@ -450,6 +486,9 @@ gst_wasapi_sink_unprepare (GstAudioSink * asink)
 {
   GstWasapiSink *self = GST_WASAPI_SINK (asink);
 
+  if (self->sharemode == AUDCLNT_SHAREMODE_EXCLUSIVE)
+    CoUninitialize ();
+
   if (self->client != NULL) {
     IAudioClient_Stop (self->client);
   }
@@ -471,33 +510,37 @@ gst_wasapi_sink_write (GstAudioSink * asink, gpointer data, guint length)
   guint pending = length;
 
   while (pending > 0) {
-    guint have_frames, can_frames, n_frames, n_frames_padding, write_len;
-
-    /* We have N frames to be written out */
-    have_frames = pending / (self->mix_format->nBlockAlign);
+    guint n_frames, write_len;
 
     WaitForSingleObject (self->event_handle, INFINITE);
 
-    /* Frames the card hasn't rendered yet */
-    hr = IAudioClient_GetCurrentPadding (self->client, &n_frames_padding);
-    if (hr != S_OK) {
-      GST_ERROR_OBJECT (self, "IAudioClient::GetCurrentPadding failed: %s",
-          gst_wasapi_util_hresult_to_string (hr));
-      length = 0;
-      goto beach;
+    if (self->sharemode == AUDCLNT_SHAREMODE_SHARED) {
+      guint have_frames, can_frames, n_frames_padding;
+
+      /* Frames the card hasn't rendered yet */
+      hr = IAudioClient_GetCurrentPadding (self->client, &n_frames_padding);
+      if (hr != S_OK) {
+        GST_ERROR_OBJECT (self, "IAudioClient::GetCurrentPadding failed: %s",
+            gst_wasapi_util_hresult_to_string (hr));
+        length = 0;
+        goto beach;
+      }
+      /* We have N frames to be written out */
+      have_frames = pending / (self->mix_format->nBlockAlign);
+      /* We can write out these many frames */
+      can_frames = self->buffer_frame_count - n_frames_padding;
+      /* We will write out these many frames, and this much length */
+      n_frames = MIN (can_frames, have_frames);
+
+      GST_TRACE_OBJECT (self, "total: %i, unread: %i, have: %i (%i bytes), "
+          "will write: %i", self->buffer_frame_count, n_frames_padding,
+          have_frames, pending, n_frames);
+    } else {
+      n_frames = self->buffer_frame_count;
     }
 
-    /* We can write out these many frames */
-    can_frames = self->buffer_frame_count - n_frames_padding;
-
-    /* We will write out these many frames, and this much length */
-    n_frames = MIN (can_frames, have_frames);
     write_len = n_frames * self->mix_format->nBlockAlign;
 
-    GST_TRACE_OBJECT (self, "total: %i, unread: %i, have: %i (%i bytes), "
-        "will write: %i (%i bytes)", self->buffer_frame_count, n_frames_padding,
-        have_frames, pending, n_frames, write_len);
-
     hr = IAudioRenderClient_GetBuffer (self->render_client, n_frames,
         (BYTE **) & dst);
     if (hr != S_OK) {
diff --git a/sys/wasapi/gstwasapisink.h b/sys/wasapi/gstwasapisink.h
index 37ba888d5..882f46e2c 100644
--- a/sys/wasapi/gstwasapisink.h
+++ b/sys/wasapi/gstwasapisink.h
@@ -40,6 +40,7 @@ struct _GstWasapiSink
 {
   GstAudioSink parent;
 
+  IMMDevice *device;
   IAudioClient *client;
   IAudioRenderClient *render_client;
   HANDLE event_handle;
@@ -57,6 +58,7 @@ struct _GstWasapiSink
 
   /* properties */
   gint role;
+  gint sharemode;
   gboolean mute;
   wchar_t *device_strid;
 };
diff --git a/sys/wasapi/gstwasapisrc.c b/sys/wasapi/gstwasapisrc.c
index 3eb19cc6c..dec93f829 100644
--- a/sys/wasapi/gstwasapisrc.c
+++ b/sys/wasapi/gstwasapisrc.c
@@ -38,8 +38,6 @@
 
 #include "gstwasapisrc.h"
 
-#include <mmdeviceapi.h>
-
 GST_DEBUG_CATEGORY_STATIC (gst_wasapi_src_debug);
 #define GST_CAT_DEFAULT gst_wasapi_src_debug
 
@@ -48,13 +46,15 @@ static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
     GST_PAD_ALWAYS,
     GST_STATIC_CAPS (GST_WASAPI_STATIC_CAPS));
 
-#define DEFAULT_ROLE    GST_WASAPI_DEVICE_ROLE_CONSOLE
+#define DEFAULT_ROLE      GST_WASAPI_DEVICE_ROLE_CONSOLE
+#define DEFAULT_EXCLUSIVE FALSE
 
 enum
 {
   PROP_0,
   PROP_ROLE,
-  PROP_DEVICE
+  PROP_DEVICE,
+  PROP_EXCLUSIVE
 };
 
 static void gst_wasapi_src_dispose (GObject * object);
@@ -108,6 +108,12 @@ gst_wasapi_src_class_init (GstWasapiSrcClass * klass)
           "WASAPI playback device as a GUID string",
           NULL, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
 
+  g_object_class_install_property (gobject_class,
+      PROP_EXCLUSIVE,
+      g_param_spec_boolean ("exclusive", "Exclusive mode",
+          "Open the device in exclusive mode",
+          DEFAULT_EXCLUSIVE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+
   gst_element_class_add_static_pad_template (gstelement_class, &src_template);
   gst_element_class_set_static_metadata (gstelement_class, "WasapiSrc",
       "Source/Audio",
@@ -206,6 +212,10 @@ gst_wasapi_src_set_property (GObject * object, guint prop_id,
           device ? g_utf8_to_utf16 (device, -1, NULL, NULL, NULL) : NULL;
       break;
     }
+    case PROP_EXCLUSIVE:
+      self->sharemode = g_value_get_boolean (value)
+          ? AUDCLNT_SHAREMODE_EXCLUSIVE : AUDCLNT_SHAREMODE_SHARED;
+      break;
     default:
       G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
       break;
@@ -226,6 +236,10 @@ gst_wasapi_src_get_property (GObject * object, guint prop_id,
       g_value_take_string (value, self->device_strid ?
           g_utf16_to_utf8 (self->device_strid, -1, NULL, NULL, NULL) : NULL);
       break;
+    case PROP_EXCLUSIVE:
+      g_value_set_boolean (value,
+          self->sharemode == AUDCLNT_SHAREMODE_EXCLUSIVE);
+      break;
     default:
       G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
       break;
@@ -238,7 +252,6 @@ gst_wasapi_src_get_caps (GstBaseSrc * bsrc, GstCaps * filter)
   GstWasapiSrc *self = GST_WASAPI_SRC (bsrc);
   WAVEFORMATEX *format = NULL;
   GstCaps *caps = NULL;
-  HRESULT hr;
 
   GST_DEBUG_OBJECT (self, "entering get caps");
 
@@ -246,16 +259,18 @@ gst_wasapi_src_get_caps (GstBaseSrc * bsrc, GstCaps * filter)
     caps = gst_caps_ref (self->cached_caps);
   } else {
     GstCaps *template_caps;
+    gboolean ret;
 
     template_caps = gst_pad_get_pad_template_caps (bsrc->srcpad);
 
     if (!self->client)
       gst_wasapi_src_open (GST_AUDIO_SRC (bsrc));
 
-    hr = IAudioClient_GetMixFormat (self->client, &format);
-    if (hr != S_OK || format == NULL) {
+    ret = gst_wasapi_util_get_device_format (GST_ELEMENT (self),
+        self->sharemode, self->device, self->client, &format);
+    if (!ret) {
       GST_ELEMENT_ERROR (self, STREAM, FORMAT, (NULL),
-          ("GetMixFormat failed: %s", gst_wasapi_util_hresult_to_string (hr)));
+          ("failed to detect format"));
       goto out;
     }
 
@@ -297,6 +312,7 @@ gst_wasapi_src_open (GstAudioSrc * asrc)
   GstWasapiSrc *self = GST_WASAPI_SRC (asrc);
   gboolean res = FALSE;
   IAudioClient *client = NULL;
+  IMMDevice *device = NULL;
 
   if (self->client)
     return TRUE;
@@ -306,7 +322,7 @@ gst_wasapi_src_open (GstAudioSrc * asrc)
    * For example, perhaps we should automatically switch to the new device if
    * the default device is changed and a device isn't explicitly selected. */
   if (!gst_wasapi_util_get_device_client (GST_ELEMENT (self), TRUE,
-          self->role, self->device_strid, &client)) {
+          self->role, self->device_strid, &device, &client)) {
     if (!self->device_strid)
       GST_ELEMENT_ERROR (self, RESOURCE, OPEN_READ, (NULL),
           ("Failed to get default device"));
@@ -317,6 +333,7 @@ gst_wasapi_src_open (GstAudioSrc * asrc)
   }
 
   self->client = client;
+  self->device = device;
   res = TRUE;
 
 beach:
@@ -329,6 +346,11 @@ gst_wasapi_src_close (GstAudioSrc * asrc)
 {
   GstWasapiSrc *self = GST_WASAPI_SRC (asrc);
 
+  if (self->device != NULL) {
+    IUnknown_Release (self->device);
+    self->device = NULL;
+  }
+
   if (self->client != NULL) {
     IUnknown_Release (self->client);
     self->client = NULL;
@@ -346,7 +368,7 @@ gst_wasapi_src_prepare (GstAudioSrc * asrc, GstAudioRingBufferSpec * spec)
   guint64 client_clock_freq = 0;
   IAudioCaptureClient *capture_client = NULL;
   REFERENCE_TIME latency_rt;
-  gint64 default_period, min_period;
+  gint64 default_period, min_period, use_period;
   guint bpf, rate, buffer_frames;
   HRESULT hr;
 
@@ -359,18 +381,30 @@ gst_wasapi_src_prepare (GstAudioSrc * asrc, GstAudioRingBufferSpec * spec)
   GST_INFO_OBJECT (self, "wasapi default period: %" G_GINT64_FORMAT
       ", min period: %" G_GINT64_FORMAT, default_period, min_period);
 
-  /* Set hnsBufferDuration to 0, which should, in theory, tell the device to
-   * create a buffer with the smallest latency possible. In practice, this is
-   * usually 2 * default_period. See:
-   * https://msdn.microsoft.com/en-us/library/windows/desktop/dd370871(v=vs.85).aspx
-   *
-   * NOTE: min_period is a lie, and I have never seen WASAPI use it as the
-   * current period */
-  hr = IAudioClient_Initialize (self->client, AUDCLNT_SHAREMODE_SHARED,
-      AUDCLNT_STREAMFLAGS_EVENTCALLBACK, 0, 0, self->mix_format, NULL);
+  if (self->sharemode == AUDCLNT_SHAREMODE_SHARED) {
+    use_period = default_period;
+    /* Set hnsBufferDuration to 0, which should, in theory, tell the device to
+     * create a buffer with the smallest latency possible. In practice, this is
+     * usually 2 * default_period. See:
+     * https://msdn.microsoft.com/en-us/library/windows/desktop/dd370871(v=vs.85).aspx
+     *
+     * NOTE: min_period is a lie, and I have never seen WASAPI use it as the
+     * current period */
+    hr = IAudioClient_Initialize (self->client, AUDCLNT_SHAREMODE_SHARED,
+        AUDCLNT_STREAMFLAGS_EVENTCALLBACK, 0, 0, self->mix_format, NULL);
+  } else {
+    use_period = default_period;
+    /* For some reason, we need to call this another time for exclusive mode */
+    CoInitialize (NULL);
+    /* FIXME: We should be able to use min_period as the device buffer size,
+     * but I'm hitting a problem in GStreamer. */
+    hr = IAudioClient_Initialize (self->client, AUDCLNT_SHAREMODE_EXCLUSIVE,
+        AUDCLNT_STREAMFLAGS_EVENTCALLBACK, use_period, use_period,
+        self->mix_format, NULL);
+  }
   if (hr != S_OK) {
     GST_ELEMENT_ERROR (self, RESOURCE, OPEN_READ, (NULL),
-        ("IAudioClient::Initialize failed: %s",
+        ("IAudioClient::Initialize () failed: %s",
             gst_wasapi_util_hresult_to_string (hr)));
     goto beach;
   }
@@ -388,8 +422,10 @@ gst_wasapi_src_prepare (GstAudioSrc * asrc, GstAudioRingBufferSpec * spec)
       "rate is %i Hz", buffer_frames, bpf, rate);
 
   spec->segsize = gst_util_uint64_scale_int_round (rate * bpf,
-      default_period * 100, GST_SECOND);
-  spec->segtotal = (buffer_frames * bpf) / spec->segsize;
+      use_period * 100, GST_SECOND);
+
+  /* We need a minimum of 2 segments to ensure glitch-free playback */
+  spec->segtotal = MAX (self->buffer_frame_count * bpf / spec->segsize, 2);
 
   GST_INFO_OBJECT (self, "segsize is %i, segtotal is %i", spec->segsize,
       spec->segtotal);
@@ -460,6 +496,9 @@ gst_wasapi_src_unprepare (GstAudioSrc * asrc)
 {
   GstWasapiSrc *self = GST_WASAPI_SRC (asrc);
 
+  if (self->sharemode == AUDCLNT_SHAREMODE_EXCLUSIVE)
+    CoUninitialize ();
+
   if (self->client != NULL) {
     IAudioClient_Stop (self->client);
   }
diff --git a/sys/wasapi/gstwasapisrc.h b/sys/wasapi/gstwasapisrc.h
index 95087bd26..fc89391a9 100644
--- a/sys/wasapi/gstwasapisrc.h
+++ b/sys/wasapi/gstwasapisrc.h
@@ -40,6 +40,7 @@ struct _GstWasapiSrc
 {
   GstAudioSrc parent;
 
+  IMMDevice *device;
   IAudioClient *client;
   IAudioClock *client_clock;
   guint64 client_clock_freq;
@@ -59,6 +60,7 @@ struct _GstWasapiSrc
 
   /* properties */
   gint role;
+  gint sharemode;
   wchar_t *device_strid;
 };
 
diff --git a/sys/wasapi/gstwasapiutil.c b/sys/wasapi/gstwasapiutil.c
index 1ad9ff6a8..9fcbd33d1 100644
--- a/sys/wasapi/gstwasapiutil.c
+++ b/sys/wasapi/gstwasapiutil.c
@@ -25,8 +25,6 @@
 #include "gstwasapiutil.h"
 #include "gstwasapidevice.h"
 
-#include <mmdeviceapi.h>
-
 /* This was only added to MinGW in ~2015 and our Cerbero toolchain is too old */
 #if defined(_MSC_VER)
 #include <functiondiscoverykeys_devpkey.h>
@@ -428,9 +426,82 @@ err:
 }
 
 gboolean
+gst_wasapi_util_get_device_format (GstElement * element,
+    gint device_mode, IMMDevice * device, IAudioClient * client,
+    WAVEFORMATEX ** ret_format)
+{
+  WAVEFORMATEX *format;
+  HRESULT hr;
+
+  *ret_format = NULL;
+
+  hr = IAudioClient_GetMixFormat (client, &format);
+  if (hr != S_OK || format == NULL) {
+    GST_ERROR_OBJECT (element, "GetMixFormat failed: %s",
+        gst_wasapi_util_hresult_to_string (hr));
+    return FALSE;
+  }
+
+  /* WASAPI always accepts the format returned by GetMixFormat in shared mode */
+  if (device_mode == AUDCLNT_SHAREMODE_SHARED)
+    goto out;
+
+  /* WASAPI may or may not support this format in exclusive mode */
+  hr = IAudioClient_IsFormatSupported (client, AUDCLNT_SHAREMODE_EXCLUSIVE,
+      format, NULL);
+  if (hr == S_OK)
+    goto out;
+
+  CoTaskMemFree (format);
+
+  /* Open the device property store, and get the format that WASAPI has been
+   * using for sending data to the device */
+  {
+    PROPVARIANT var;
+    IPropertyStore *prop_store = NULL;
+
+    hr = IMMDevice_OpenPropertyStore (device, STGM_READ, &prop_store);
+    if (hr != S_OK) {
+      GST_ERROR_OBJECT (element, "OpenPropertyStore failed: %s",
+          gst_wasapi_util_hresult_to_string (hr));
+      return FALSE;
+    }
+
+    hr = IPropertyStore_GetValue (prop_store, &PKEY_AudioEngine_DeviceFormat,
+        &var);
+    if (hr != S_OK) {
+      GST_ERROR_OBJECT (element, "GetValue failed: %s",
+          gst_wasapi_util_hresult_to_string (hr));
+      IUnknown_Release (prop_store);
+      return FALSE;
+    }
+
+    format = malloc (var.blob.cbSize);
+    memcpy (format, var.blob.pBlobData, var.blob.cbSize);
+
+    PropVariantClear (&var);
+    IUnknown_Release (prop_store);
+  }
+
+  /* WASAPI may or may not support this format in exclusive mode */
+  hr = IAudioClient_IsFormatSupported (client, AUDCLNT_SHAREMODE_EXCLUSIVE,
+      format, NULL);
+  if (hr == S_OK)
+    goto out;
+
+  GST_ERROR_OBJECT (element, "AudioEngine DeviceFormat not supported");
+  free (format);
+  return FALSE;
+
+out:
+  *ret_format = format;
+  return TRUE;
+}
+
+gboolean
 gst_wasapi_util_get_device_client (GstElement * element,
     gboolean capture, gint role, const wchar_t * device_strid,
-    IAudioClient ** ret_client)
+    IMMDevice ** ret_device, IAudioClient ** ret_client)
 {
   gboolean res = FALSE;
   HRESULT hr;
@@ -468,7 +539,9 @@ gst_wasapi_util_get_device_client (GstElement * element,
   }
 
   IUnknown_AddRef (client);
+  IUnknown_AddRef (device);
   *ret_client = client;
+  *ret_device = device;
 
   res = TRUE;
 
diff --git a/sys/wasapi/gstwasapiutil.h b/sys/wasapi/gstwasapiutil.h
index af44895a6..b8337de1a 100644
--- a/sys/wasapi/gstwasapiutil.h
+++ b/sys/wasapi/gstwasapiutil.h
@@ -25,6 +25,7 @@
 #include <gst/audio/gstaudiosrc.h>
 #include <gst/audio/gstaudiosink.h>
 
+#include <mmdeviceapi.h>
 #include <audioclient.h>
 
 /* Static Caps shared between source, sink, and device provider */
@@ -57,7 +58,11 @@ gboolean gst_wasapi_util_get_devices (GstElement * element, gboolean active,
 
 gboolean gst_wasapi_util_get_device_client (GstElement * element,
     gboolean capture, gint role, const wchar_t * device_strid,
-    IAudioClient ** ret_client);
+    IMMDevice ** ret_device, IAudioClient ** ret_client);
+
+gboolean gst_wasapi_util_get_device_format (GstElement * element,
+    gint device_mode, IMMDevice * device, IAudioClient * client,
+    WAVEFORMATEX ** ret_format);
 
 gboolean gst_wasapi_util_get_render_client (GstElement * element,
     IAudioClient * client, IAudioRenderClient ** ret_render_client);