summaryrefslogtreecommitdiffstats
path: root/Whisper/API/MfStructs.h
diff options
context:
space:
mode:
authorKonstantin <const@const.me>2023-01-16 14:52:43 +0100
committerKonstantin <const@const.me>2023-01-16 14:52:43 +0100
commit8c4603c73675958efc960fbd4bb599a2909d106a (patch)
tree714dc6fc9a1672d5fd7f89676b97e10959662abc /Whisper/API/MfStructs.h
parent990a8d0dbaefc996244097397259e92758b15cce (diff)
Source codes
Diffstat (limited to 'Whisper/API/MfStructs.h')
-rw-r--r--Whisper/API/MfStructs.h51
1 files changed, 51 insertions, 0 deletions
diff --git a/Whisper/API/MfStructs.h b/Whisper/API/MfStructs.h
new file mode 100644
index 0000000..cd27659
--- /dev/null
+++ b/Whisper/API/MfStructs.h
@@ -0,0 +1,51 @@
+#pragma once
+
+namespace Whisper
+{
+ struct sCaptureDevice
+ {
+ // The display name is suitable for showing to the user, but might not be unique.
+ const wchar_t* displayName;
+
+ // Endpoint ID for an audio capture device
+ // It uniquely identifies the device on the system, but is not a readable string.
+ const wchar_t* endpoint;
+ };
+
+ using pfnFoundCaptureDevices = HRESULT( __stdcall* )( int len, const sCaptureDevice* buffer, void* pv );
+
+ // Flags for the audio capture
+ enum struct eCaptureFlags : uint32_t
+ {
+ // When the capture device supports stereo, keep stereo PCM samples in addition to mono
+ Stereo = 1,
+ };
+
+ // Parameters for audio capture
+ struct sCaptureParams
+ {
+ float minDuration = 2.0f;
+ float maxDuration = 3.0f;
+ float dropStartSilence = 0.25f;
+ float pauseDuration = 0.333f;
+ // Flags for the audio capture
+ uint32_t flags = 0;
+ };
+
+ enum struct eCaptureStatus : uint8_t
+ {
+ Listening = 1,
+ Voice = 2,
+ Transcribing = 4,
+ Stalled = 0x80,
+ };
+
+ using pfnShouldCancel = HRESULT( __stdcall* )( void* pv ) noexcept;
+ using pfnCaptureStatus = HRESULT( __stdcall* )( void* pv, eCaptureStatus status ) noexcept;
+ struct sCaptureCallbacks
+ {
+ pfnShouldCancel shouldCancel;
+ pfnCaptureStatus captureStatus;
+ void* pv;
+ };
+} \ No newline at end of file