DLL API for diarize feature

5 files changed, 19 insertions, 0 deletions

diff --git a/Whisper/API/TranscribeStructs.h b/Whisper/API/TranscribeStructs.h
index ac28357..29bb9ab 100644
--- a/Whisper/API/TranscribeStructs.h
+++ b/Whisper/API/TranscribeStructs.h
@@ -124,4 +124,13 @@ namespace Whisper
 	{
 		return 0 != ( (uint32_t)a & (uint32_t)b );
 	}
+
+	// Output value for iContext.detectSpeaker method
+	enum struct eSpeakerChannel : uint8_t
+	{
+		Unsure = 0,
+		Left = 1,
+		Right = 2,
+		NoStereoData = 0xFF,
+	};
 }
\ No newline at end of file
diff --git a/Whisper/API/iContext.cl.h b/Whisper/API/iContext.cl.h
index fdb15ce..18e56de 100644
--- a/Whisper/API/iContext.cl.h
+++ b/Whisper/API/iContext.cl.h
@@ -31,6 +31,8 @@ namespace Whisper
 		virtual HRESULT COMLIGHTCALL runCapture( const sFullParams& params, const sCaptureCallbacks& callbacks, const iAudioCapture* reader ) = 0;
 
 		virtual HRESULT COMLIGHTCALL getResults( eResultFlags flags, iTranscribeResult** pp ) const = 0;
+		// Try to detect speaker by comparing channels of the stereo PCM data
+		virtual HRESULT COMLIGHTCALL detectSpeaker( const sTimeInterval& time, eSpeakerChannel& result ) const = 0;
 
 		virtual HRESULT COMLIGHTCALL getModel( iModel** pp ) = 0;
 
diff --git a/Whisper/API/iContext.h b/Whisper/API/iContext.h
index d6ca29c..fc38a53 100644
--- a/Whisper/API/iContext.h
+++ b/Whisper/API/iContext.h
@@ -28,6 +28,8 @@ namespace Whisper
 		HRESULT __stdcall runCapture( const sFullParams& params, const sCaptureCallbacks& callbacks, const iAudioCapture* reader );
 
 		HRESULT __stdcall getResults( eResultFlags flags, iTranscribeResult** pp ) const;
+		// Try to detect speaker by comparing channels of the stereo PCM data
+		HRESULT __stdcall detectSpeaker( const sTimeInterval& time, eSpeakerChannel& result ) const;
 
 		HRESULT __stdcall getModel( iModel** pp );
 
diff --git a/Whisper/Whisper/ContextImpl.h b/Whisper/Whisper/ContextImpl.h
index 971f629..448efd5 100644
--- a/Whisper/Whisper/ContextImpl.h
+++ b/Whisper/Whisper/ContextImpl.h
@@ -63,6 +63,7 @@ namespace Whisper
 		HRESULT COMLIGHTCALL makeResults( eResultFlags flags, TranscribeResult& res ) const noexcept;
 
 		HRESULT COMLIGHTCALL getResults( eResultFlags flags, iTranscribeResult** pp ) const noexcept override final;
+		HRESULT COMLIGHTCALL detectSpeaker( const sTimeInterval& time, eSpeakerChannel& result ) const noexcept override final;
 
 		int defaultThreadsCount() const;
 
diff --git a/Whisper/Whisper/ContextImpl.misc.cpp b/Whisper/Whisper/ContextImpl.misc.cpp
index 9a156fb..9eb4c04 100644
--- a/Whisper/Whisper/ContextImpl.misc.cpp
+++ b/Whisper/Whisper/ContextImpl.misc.cpp
@@ -401,4 +401,9 @@ HRESULT COMLIGHTCALL ContextImpl::runStreamed( const sFullParams& params, const
 	{
 		return hr;
 	}
+}
+
+HRESULT COMLIGHTCALL ContextImpl::detectSpeaker( const sTimeInterval& time, eSpeakerChannel& result ) const noexcept
+{
+	return E_NOTIMPL;
 }
\ No newline at end of file