TTS implemented using correct resampling from 48kHz to 44.1kHz

CMakeLists.txt

@@ -316,6 +316,7 @@ add_definitions( "-DVACORE_CMAKE_DATE=\"${VACORE_CMAKE_DATE}\"" )
 
 if( ITA_VACORE_WITH_TTS_SIGNAL_SOURCE )
 	vista_use_package( CereVoice REQUIRED FIND_DEPENDENCIES)
+	vista_use_package( samplerate REQUIRED FIND_DEPENDENCIES)
 	if( VCEREVOICE_FOUND )
 		add_definitions( "-DVACORE_WITH_TTS_SIGNAL_SOURCE" )
 	endif( )

src/Audiosignals/VATextToSpeechSignalSource.cpp

@@ -19,10 +19,13 @@
 
 #include <VistaTools/VistaFileSystemDirectory.h>
 
+
+
 #ifdef VACORE_WITH_TTS_SIGNAL_SOURCE
 #include <cerevoice_eng.h>
 #endif
 
+//#define TTS_USE_WAV
 
 CVATextToSpeechSignalSource::CVATextToSpeechSignalSource( const double dSampleRate, const int iBlockLength )
 : ITADatasourceRealization(1, dSampleRate, (unsigned int)(iBlockLength))
@@ -94,7 +97,6 @@ const float* CVATextToSpeechSignalSource::GetStreamBlock( const CVAAudiostreamSt
 		//	return pAudioFile->GetStreamBlock(pStreamInfo);
 	}
 
-	// @todo: take samples from generated WAV file
 
 	return m_sbOut.data();
 }
@@ -209,14 +211,14 @@ void CVATextToSpeechSignalSource::SetParameters( const CVAStruct& oParams )
 		if (oParams.HasKey("voice") && oParams["voice"].IsString())
 			sVoice = oParams["voice"];
 
-		bool direct_replay = false;
-		if (oParams.HasKey("direct_replay") && oParams["direct_replay"].IsBool() && oParams["direct_replay"]){
-			direct_replay = true;
+		bool direct_playback = false;
+		if (oParams.HasKey("direct_playback") && oParams["direct_playback"].IsBool() && oParams["direct_playback"]){
+			direct_playback = true;
 		}
 
 		std::string id = "tmp";
 		if (!oParams.HasKey("id") || !oParams["id"].IsString()){
-			if (!direct_replay){
+			if (!direct_playback){
 				VA_WARN("CVATextToSpeechSignalSource", "No id is given for the prepare speech request, the user application has to give an unique id.");
 				return;
 			}
@@ -242,23 +244,35 @@ void CVATextToSpeechSignalSource::SetParameters( const CVAStruct& oParams )
 		CPRCEN_engine_channel_reset(TTSEngine::getInstance().getEngine(), chan);
 		CPRCEN_engine_clear_callback(TTSEngine::getInstance().getEngine(), chan);
 		CPRCEN_engine_set_callback(TTSEngine::getInstance().getEngine(), chan, (void*)&data, VisemeProcessing);
-		//CPRCEN_engine_channel_to_file(TTSEngine::getInstance().getEngine(), chan, "D:/work/tts.wav", CPRCEN_RIFF); /* File output on channel */
+#ifdef TTS_USE_WAV
+		CPRCEN_engine_channel_to_file(TTSEngine::getInstance().getEngine(), chan, "D:/work/tts.wav", CPRCEN_RIFF); /* File output on channel */
+#endif
 		CPRC_abuf* buf = CPRCEN_engine_channel_speak(TTSEngine::getInstance().getEngine(), chan, sText.c_str(), sText.length(), true);
-		int sRate = CPRC_abuf_wav_srate(buf);
-
-		std::cout << "CereVoice rate is: " << sRate << " while sound signal rate is: " << GetSampleRate() << std::endl;
 
 		data.visemes += "</speech>\n";
-		data.visemes += "<event id=\"" + id + "\" start=\"0\" message=\"speech started\"  />";
+		//data.visemes += "<event id=\"" + id + "\" start=\"0\" message=\"speech started\"  />";
 
 		m_Visemes[id] = data.visemes;
 
+#ifdef TTS_USE_WAV
+		ITASampleFrame* pAudioBuffer = new  ITASampleFrame("D:/work/tts.wav");	
+
+		CITAAudioSample* pAudioSample = new CITAAudioSample();
+		//pAudioSample->Load(*pAudioBuffer, 44100.0f);
+		pAudioSample->Load(*pAudioBuffer, TTSEngine::getInstance().getSampleRate());
+		m_AudioSampleFrames[id] = pAudioSample;
+#else
 		ITASampleFrame* pAudioBuffer = new  ITASampleFrame();
 		pAudioBuffer->init(1, data.floatBuffer.size(), false);
 		(*pAudioBuffer)[0].write(&data.floatBuffer[0], data.floatBuffer.size());
-		m_AudioSampleFrames[id] = pAudioBuffer;
 
-		if (direct_replay){
+		CITAAudioSample* pAudioSample = new CITAAudioSample();
+		pAudioSample->Load(*pAudioBuffer, TTSEngine::getInstance().getSampleRate());
+		m_AudioSampleFrames[id] = pAudioSample;
+#endif
+		
+
+		if (direct_playback){
 			CVAStruct oParams_play;
 			oParams_play["play_speech"] = id;
 			oParams_play["free_after"] = true;
@@ -305,6 +319,8 @@ void CVATextToSpeechSignalSource::SetParameters( const CVAStruct& oParams )
 
 		ITASampleFrame* pAudioBuffer = it->second;
 
+		VA_INFO("CVATextToSpeechSignalSource", "GetSAmpleRAte: " + std::to_string(GetSampleRate()) + "  engine::rate " + std::to_string(TTSEngine::getInstance().getSampleRate()));
+
 		if (m_pBufferDataSource != NULL)
 			delete m_pBufferDataSource;
 		m_pBufferDataSource = new ITABufferDatasource((*pAudioBuffer)[0].data(), pAudioBuffer->length(), GetSampleRate(), GetBlocklength());
@@ -426,6 +442,18 @@ CVATextToSpeechSignalSource::TTSEngine::TTSEngine(){
 		VA_INFO("CVATextToSpeechSignalSource", "Loaded voice \"" + name + "\"");
 	}
 
+	//now check for the sampled voices which sample rate we have
+	m_sampleRate = -1.0;
+	int num_voices = CPRCEN_engine_get_voice_count(getEngine());
+	for (int i = 0; i < num_voices; i++) {
+		std::string strSamplerate = CPRCEN_engine_get_voice_info(getEngine(), i, "SAMPLE_RATE");
+		float rate = std::stof(strSamplerate);
+		if (m_sampleRate < 0.0)
+			m_sampleRate = rate;
+		if (rate != m_sampleRate)
+			VA_WARN("CVATextToSpeechSignalSource", "Voices with different sample rates are used namely "+std::to_string(rate)+" and "+std::to_string(m_sampleRate));
+	}
+
 	SetupPhonemeMapping();
 
 	//std::string licence_file = voices_path + "cerevoice_heather_4.0.0_48k.lic";
@@ -447,6 +475,10 @@ CPRCEN_engine* CVATextToSpeechSignalSource::TTSEngine::getEngine() const
 	return m_pTTSEngine;
 }
 
+float CVATextToSpeechSignalSource::TTSEngine::getSampleRate() const {
+	return m_sampleRate;
+}
+
 std::string CVATextToSpeechSignalSource::TTSEngine::PhonemeToViseme(std::string phoneme) {
 
 	auto it = m_phonemeToId.find(phoneme);

src/Audiosignals/VATextToSpeechSignalSource.h

@@ -8,6 +8,7 @@
 
 #include <ITADataSourceRealization.h>
 #include <ITASampleBuffer.h>
+#include <ITAAudioSample.h>
 #include <ITAAtomicPrimitives.h>
 
 class ITABufferDatasource;
@@ -15,6 +16,7 @@ class CVACoreImpl;
 class CPRCEN_engine;
 class CPRC_abuf;
 
+
 /** Text-to-speech signal source
   * 
   * The TTS signal source generates sound from text using external libraries, like TTSRelay for Windows platforms.
@@ -41,7 +43,7 @@ public:
 	*		["prepare_text"] = text to be spoken
 	*		["id"] = identificator that will be used for playing this speech and reference it (must be unique)
 	*		["voice"] = the voice to be used //if none is given or the one given cannot be found the standard voice is used (i.e. "Heather")
-	*		["direct_replay"] = true/false whether the audio should directly be played (in this case no id has to be given, should not be used for lipsyncing) 
+	*		["direct_playback"] = true/false whether the audio should directly be played (in this case no id has to be given, should not be used for lipsyncing) 
 
 	*	This should be used to start a prepared TTS using a CVAStruct with:
 	*		["play_speech"] = identificator (int) of created speech	
@@ -86,6 +88,7 @@ private:
 		void operator=(TTSEngine const&) = delete;
 
 		CPRCEN_engine* getEngine() const;
+		float getSampleRate() const;
 		std::string PhonemeToViseme(std::string phoneme);
 
 	private:
@@ -94,6 +97,7 @@ private:
 		CPRCEN_engine* m_pTTSEngine; //you must not delete this from outside!!!!!!
 		std::map<std::string, int> m_phonemeToId;
 		std::map<int, std::string> m_idToViseme;
+		float m_sampleRate;
 	};
 
 	struct UserCallbackData{
@@ -105,7 +109,7 @@ private:
 	static void VisemeProcessing(CPRC_abuf* abuf, void * userdata); //used as callback for the CereVoice engine
 	static std::string to_string_with_precision(float a_value, const int n = 3);
 
-	std::map<std::string, ITASampleFrame*> m_AudioSampleFrames;
+	std::map<std::string, CITAAudioSample*> m_AudioSampleFrames;
 	std::map<std::string, std::string> m_Visemes;
 
 	IVACore* m_pAssociatedCore;