From e1e3ac09a97d602a6ea60ff1928de77de81d99a7 Mon Sep 17 00:00:00 2001
From: Konstantin <const@const.me>
Date: Sat, 28 Jan 2023 15:18:02 +0100
Subject: DLL API for diarize feature

---
 Whisper/API/TranscribeStructs.h      |  9 +++++++++
 Whisper/API/iContext.cl.h            |  2 ++
 Whisper/API/iContext.h               |  2 ++
 Whisper/Whisper/ContextImpl.h        |  1 +
 Whisper/Whisper/ContextImpl.misc.cpp |  5 +++++
 WhisperNet/API/eSpeakerChannel.cs    | 15 +++++++++++++++
 WhisperNet/Internal/iContext.cs      |  4 ++++
 7 files changed, 38 insertions(+)
 create mode 100644 WhisperNet/API/eSpeakerChannel.cs

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
diff --git a/WhisperNet/API/eSpeakerChannel.cs b/WhisperNet/API/eSpeakerChannel.cs
new file mode 100644
index 0000000..edb96e0
--- /dev/null
+++ b/WhisperNet/API/eSpeakerChannel.cs
@@ -0,0 +1,15 @@
+namespace Whisper
+{
+	/// <summary>Output value for iContext.detectSpeaker method</summary>
+	public enum eSpeakerChannel: byte
+	{
+		/// <summary>Unable to detect</summary>
+		Unsure = 0,
+		/// <summary>The speech was mostly in the left channel</summary>
+		Left = 1,
+		/// <summary>The speech was mostly in the right channel</summary>
+		Right = 2,
+		/// <summary>The audio only has 1 channel</summary>
+		NoStereoData = 0xFF,
+	}
+}
\ No newline at end of file
diff --git a/WhisperNet/Internal/iContext.cs b/WhisperNet/Internal/iContext.cs
index 6adf8c5..010c139 100644
--- a/WhisperNet/Internal/iContext.cs
+++ b/WhisperNet/Internal/iContext.cs
@@ -21,6 +21,10 @@ namespace Whisper.Internal
 		[RetValIndex( 1 )]
 		iTranscribeResult getResults( eResultFlags flags );
 
+		/// <summary>Try to detect speaker by comparing channels of the stereo PCM data</summary>
+		[RetValIndex( 1 )]
+		eSpeakerChannel detectSpeaker( [In] ref sTimeInterval interval );
+
 		/// <summary>Get the model which was used to create this context</summary>
 		[RetValIndex]
 		iModel getModel();
-- 
cgit v1.2.3