Win-specific: Add volume and rate tuning for TTS, fix some errors

articleview.cc

@@ -973,7 +973,7 @@ void ArticleView::openLink( QUrl const & url, QUrl const & ref,
 
       if ( itemMd5Id == md5Id )
       {
-        SpeechClient * speechClient = new SpeechClient( i->id, this );
+        SpeechClient * speechClient = new SpeechClient( *i, this );
         connect( speechClient, SIGNAL( finished() ), speechClient, SLOT( deleteLater() ) );
         speechClient->tell( text );
         break;

config.cc

@@ -604,7 +604,7 @@ Class load() throw( exError )
 
   QDomNode ves = root.namedItem( "voiceEngines" );
 
-	if ( !wss.isNull() )
+  if ( !ves.isNull() )
   {
     QDomNodeList nl = ves.toElement().elementsByTagName( "voiceEngine" );
 
@@ -617,6 +617,12 @@ Class load() throw( exError )
       v.id = ve.attribute( "id" );
       v.name = ve.attribute( "name" );
       v.iconFilename = ve.attribute( "icon" );
+      v.volume = ve.attribute( "volume", "50" ).toInt();
+      if( v.volume < 0 || v.volume > 100 )
+        v.volume = 50;
+      v.rate = ve.attribute( "rate", "50" ).toInt();
+      if( v.rate < 0 || v.rate > 100 )
+        v.rate = 50;
       c.voiceEngines.push_back( v );
     }
   }
@@ -1205,6 +1211,14 @@ void save( Class const & c ) throw( exError )
       QDomAttr icon = dd.createAttribute( "icon" );
       icon.setValue( i->iconFilename );
       v.setAttributeNode( icon );
+
+      QDomAttr volume = dd.createAttribute( "volume" );
+      volume.setValue( QString::number( i->volume ) );
+      v.setAttributeNode( volume );
+
+      QDomAttr rate = dd.createAttribute( "rate" );
+      rate.setValue( QString::number( i->rate ) );
+      v.setAttributeNode( rate );
     }
   }
 

config.hh

@@ -387,17 +387,29 @@ struct VoiceEngine
   QString id;
   QString name;
   QString iconFilename;
+  int volume; // 0-100 allowed
+  int rate;   // 0-100 allowed
 
   VoiceEngine(): enabled( false )
-  {
-  }
+    , volume( 50 )
+    , rate( 50 )
+  {}
+  VoiceEngine( QString id_, QString name_, int volume_, int rate_ ):
+    enabled( false )
+    , id( id_ )
+    , name( name_ )
+    , volume( volume_ )
+    , rate( rate_ )
+  {}
 
   bool operator == ( VoiceEngine const & other ) const
   {
     return enabled == other.enabled &&
            id == other.id &&
            name == other.name &&
-           iconFilename == other.iconFilename;
+           iconFilename == other.iconFilename &&
+           volume == other.volume &&
+           rate == other.rate;
   }
 
   bool operator != ( VoiceEngine const & other ) const

goldendict.pro

@@ -348,7 +348,8 @@ win32 {
                texttospeechsource.hh \
                sapi.hh \
                sphelper.hh \
-               speechclient.hh
+               speechclient.hh \
+               speechhlp.hh
 }
 
 mac {

speechclient.hh

@@ -2,6 +2,7 @@
 #define __SPEECHCLIENT_HH_INCLUDED__
 
 #include <QObject>
+#include "config.hh"
 
 class SpeechClient: public QObject
 {
@@ -13,18 +14,28 @@ public:
   {
     QString id;
     QString name;
+    // Volume and rate may vary from 0 to 100
+    int volume;
+    int rate;
+    Engine( Config::VoiceEngine const & e ) :
+      id( e.id )
+      , name( e.name )
+      , volume( e.volume )
+      , rate( e.rate )
+    {}
   };
 
   typedef QList<Engine> Engines;
 
-  SpeechClient( QString const & engineId, QObject * parent = 0L );
+  SpeechClient( Config::VoiceEngine const & e, QObject * parent = 0L );
   virtual ~SpeechClient();
 
   static Engines availableEngines();
 
   const Engine & engine() const;
-  bool tell( QString const & text );
-  bool say( QString const & text );
+
+  bool tell( QString const & text, int volume = -1, int rate = -1 );
+  bool say( QString const & text, int volume = -1, int rate = -1 );
 
 signals:
   void started( bool ok );

speechclient_win.cc

@@ -7,10 +7,13 @@
 
 struct SpeechClient::InternalData
 {
-  InternalData( QString const & engineId ):
+  InternalData( Config::VoiceEngine const & e ):
     waitingFinish( false )
+    , engine( e )
+    , oldVolume( -1 )
+    , oldRate( -1 )
   {
-    sp = speechCreate( engineId.toStdWString().c_str() );
+    sp = speechCreate( e.id.toStdWString().c_str() );
   }
 
   ~InternalData()
@@ -19,13 +22,15 @@ struct SpeechClient::InternalData
   }
 
   SpeechHelper sp;
-  Engine engine;
   bool waitingFinish;
+  Engine engine;
+  int oldVolume;
+  int oldRate;
 };
 
-SpeechClient::SpeechClient( QString const & engineId, QObject * parent ):
+SpeechClient::SpeechClient( Config::VoiceEngine const & e, QObject * parent ):
   QObject( parent ),
-  internalData( new InternalData( engineId ) )
+  internalData( new InternalData( e ) )
 {
 }
 
@@ -40,11 +45,10 @@ static bool enumEngines( void * /* token */,
                          void * userData )
 {
   SpeechClient::Engines * pEngines = ( SpeechClient::Engines * )userData;
-  SpeechClient::Engine engine =
-  {
+  SpeechClient::Engine engine( Config::VoiceEngine(
     QString::fromWCharArray( id ),
-    QString::fromWCharArray( name )
-  };
+    QString::fromWCharArray( name ),
+    50, 50 ) );
   pEngines->push_back( engine );
   return true;
 }
@@ -61,7 +65,7 @@ const SpeechClient::Engine & SpeechClient::engine() const
   return internalData->engine;
 }
 
-bool SpeechClient::tell( QString const & text )
+bool SpeechClient::tell( QString const & text, int volume, int rate )
 {
   if ( !speechAvailable( internalData->sp ) )
     return false;
@@ -69,7 +73,16 @@ bool SpeechClient::tell( QString const & text )
   if ( internalData->waitingFinish )
     return false;
 
+  if( volume < 0 )
+    volume = engine().volume;
+  if( rate < 0 )
+    rate = engine().rate;
+
+  internalData->oldVolume = setSpeechVolume( internalData->sp, volume );
+  internalData->oldRate = setSpeechRate( internalData->sp, rate );
+
   bool ok = speechTell( internalData->sp, text.toStdWString().c_str() );
+
   emit started( ok );
 
   if ( ok )
@@ -84,12 +97,27 @@ bool SpeechClient::tell( QString const & text )
   return ok;
 }
 
-bool SpeechClient::say( QString const & text )
+bool SpeechClient::say( QString const & text, int volume, int rate )
 {
   if ( !speechAvailable( internalData->sp ) )
     return false;
 
-  return speechSay( internalData->sp, text.toStdWString().c_str() );
+  if( volume < 0 )
+    volume = engine().volume;
+  if( rate < 0 )
+    rate = engine().rate;
+
+  int oldVolume = setSpeechVolume( internalData->sp, volume );
+  int oldRate = setSpeechRate( internalData->sp, rate );
+
+  bool ok = speechSay( internalData->sp, text.toStdWString().c_str() );
+
+  if( oldVolume >=0 )
+    setSpeechVolume( internalData->sp, oldVolume );
+  if( oldRate >=0 )
+    setSpeechRate( internalData->sp, oldRate );
+
+  return ok;
 }
 
 void SpeechClient::timerEvent( QTimerEvent * evt )
@@ -103,6 +131,14 @@ void SpeechClient::timerEvent( QTimerEvent * evt )
   {
     killTimer( evt->timerId() ) ;
     internalData->waitingFinish = false;
+
+    if( internalData->oldVolume >=0 )
+      setSpeechVolume( internalData->sp, internalData->oldVolume );
+    if( internalData->oldRate >=0 )
+      setSpeechRate( internalData->sp, internalData->oldRate );
+    internalData->oldVolume = -1;
+    internalData->oldRate = -1;
+
     emit finished();
   }
 }

speechhlp.cc

@@ -165,3 +165,27 @@ bool speechSay(SpeechHelper sp, const wchar_t *text)
     HRESULT hr = sp->voice->Speak(text, SPF_IS_NOT_XML, 0);
     return !!SUCCEEDED(hr);
 }
+
+int setSpeechVolume( SpeechHelper sp, int newVolume )
+{
+  if( !sp || !sp->voice || newVolume < 0 || newVolume > 100 )
+    return -1;
+  unsigned short oldVolume;
+  HRESULT hr = sp->voice->GetVolume( &oldVolume );
+  if( !SUCCEEDED( hr ) )
+    return -1;
+  sp->voice->SetVolume( (unsigned short) newVolume );
+  return oldVolume;
+}
+
+int setSpeechRate( SpeechHelper sp, int newRate )
+{
+  if( !sp || !sp->voice || newRate < 0 || newRate > 100 )
+    return -1;
+  long oldRate;
+  HRESULT hr = sp->voice->GetRate( &oldRate );
+  if( !SUCCEEDED( hr ) )
+    return -1;
+  sp->voice->SetRate( ( newRate - 50 ) / 5 );
+  return oldRate * 5 + 50;
+}

speechhlp.hh

@@ -17,6 +17,8 @@ const wchar_t * speechEngineName(SpeechHelper sp);
 bool speechTell(SpeechHelper sp, const wchar_t *text);
 bool speechTellFinished(SpeechHelper sp);
 bool speechSay(SpeechHelper sp, const wchar_t *text);
+int setSpeechVolume( SpeechHelper sp, int newVolume );
+int setSpeechRate( SpeechHelper sp, int newRate );
 
 #ifdef __cplusplus
 }

sphelper.hh

@@ -17,6 +17,11 @@
 #define SPERR_NOT_FOUND                MAKE_SAPI_ERROR(0x03a)
 #endif
 
+#ifdef _SAPI_VER
+#undef _SAPI_VER
+#endif
+#define _SAPI_VER 0503
+
 inline void SpHexFromUlong(WCHAR * psz, ULONG ul)
 {
     // If for some reason we cannot convert a number, set it to 0

texttospeechsource.cc

@@ -25,6 +25,30 @@ TextToSpeechSource::TextToSpeechSource( QWidget * parent,
   {
     ui.availableVoiceEngines->addItem( engine.name, engine.id );
   }
+
+  if( voiceEngines.count() > 0 )
+  {
+    QModelIndex const &idx = ui.selectedVoiceEngines->model()->index( 0, 0 );
+    if( idx.isValid() )
+      ui.selectedVoiceEngines->setCurrentIndex( idx );
+  }
+
+  adjustSliders();
+
+  connect( ui.volumeSlider, SIGNAL( valueChanged( int ) ),
+           this, SLOT( slidersChanged() ) );
+  connect( ui.rateSlider, SIGNAL( valueChanged( int ) ),
+           this, SLOT( slidersChanged() ) );
+  connect( ui.selectedVoiceEngines->selectionModel(), SIGNAL( selectionChanged( QItemSelection, QItemSelection ) ),
+           this, SLOT( selectionChanged() ) );
+}
+
+void TextToSpeechSource::slidersChanged()
+{
+  if( ui.selectedVoiceEngines->selectionModel()->hasSelection() )
+    voiceEnginesModel.setEngineParams( ui.selectedVoiceEngines->currentIndex(),
+                                       ui.volumeSlider->value(),
+                                       ui.rateSlider->value()) ;
 }
 
 void TextToSpeechSource::on_addVoiceEngine_clicked()
@@ -37,12 +61,15 @@ void TextToSpeechSource::on_addVoiceEngine_clicked()
     return;
   }
 
-  // Fake id and name
-  QString name = ui.availableVoiceEngines->itemText( 0 );
-  QString id = ui.availableVoiceEngines->itemData( 0 ).toString();
-  voiceEnginesModel.addNewVoiceEngine( id, name );
+  int idx = ui.availableVoiceEngines->currentIndex();
+  if( idx >= 0 )
+  {
+    QString name = ui.availableVoiceEngines->itemText( idx );
+    QString id = ui.availableVoiceEngines->itemData( idx ).toString();
+    voiceEnginesModel.addNewVoiceEngine( id, name, ui.volumeSlider->value(), ui.rateSlider->value() );
     fitSelectedVoiceEnginesColumns();
   }
+}
 
 void TextToSpeechSource::on_removeVoiceEngine_clicked()
 {
@@ -66,8 +93,12 @@ void TextToSpeechSource::on_previewVoice_clicked()
     return;
 
   QString engineId = ui.availableVoiceEngines->itemData( idx ).toString();
+  QString name = ui.availableVoiceEngines->itemText( idx );
   QString text = ui.previewText->text();
-  SpeechClient * speechClient = new SpeechClient( engineId, this );
+  int volume = ui.volumeSlider->value();
+  int rate = ui.rateSlider->value();
+
+  SpeechClient * speechClient = new SpeechClient( Config::VoiceEngine( engineId, name, volume, rate ), this );
 
   connect( speechClient, SIGNAL( started( bool ) ), ui.previewVoice, SLOT( setDisabled( bool ) ) );
   connect( speechClient, SIGNAL( finished() ), this, SLOT( previewVoiceFinished() ) );
@@ -87,6 +118,35 @@ void TextToSpeechSource::fitSelectedVoiceEnginesColumns()
   ui.selectedVoiceEngines->resizeColumnToContents( VoiceEnginesModel::kColumnIcon );
 }
 
+void TextToSpeechSource::adjustSliders()
+{
+  QModelIndex const & index = ui.selectedVoiceEngines->currentIndex();
+  if ( index.isValid() )
+  {
+    Config::VoiceEngines const &engines = voiceEnginesModel.getCurrentVoiceEngines();
+    ui.volumeSlider->setValue( engines[ index.row() ].volume );
+    ui.rateSlider->setValue( engines[ index.row() ].rate );
+    return;
+  }
+  ui.volumeSlider->setValue( 50 );
+  ui.rateSlider->setValue( 50 );
+}
+
+void TextToSpeechSource::selectionChanged()
+{
+  disconnect( ui.volumeSlider, SIGNAL( valueChanged( int ) ),
+              this, SLOT( slidersChanged() ) );
+  disconnect( ui.rateSlider, SIGNAL( valueChanged( int ) ),
+              this, SLOT( slidersChanged() ) );
+
+  adjustSliders();
+
+  connect( ui.volumeSlider, SIGNAL( valueChanged( int ) ),
+           this, SLOT( slidersChanged() ) );
+  connect( ui.rateSlider, SIGNAL( valueChanged( int ) ),
+           this, SLOT( slidersChanged() ) );
+}
+
 VoiceEnginesModel::VoiceEnginesModel( QWidget * parent,
                                       Config::VoiceEngines const & voiceEngines ):
   QAbstractItemModel( parent ), voiceEngines( voiceEngines )
@@ -100,7 +160,8 @@ void VoiceEnginesModel::removeVoiceEngine( int index )
   endRemoveRows();
 }
 
-void VoiceEnginesModel::addNewVoiceEngine( QString const & id, QString const & name )
+void VoiceEnginesModel::addNewVoiceEngine( QString const & id, QString const & name,
+                                           int volume, int rate )
 {
   if ( id.isEmpty() || name.isEmpty() )
     return;
@@ -109,6 +170,8 @@ void VoiceEnginesModel::addNewVoiceEngine( QString const & id, QString const & n
   v.enabled = true;
   v.id = id;
   v.name = name;
+  v.volume = volume;
+  v.rate = rate;
 
   beginInsertRows( QModelIndex(), voiceEngines.size(), voiceEngines.size() );
   voiceEngines.push_back( v );
@@ -243,6 +306,15 @@ bool VoiceEnginesModel::setData( QModelIndex const & index, const QVariant & val
   return false;
 }
 
+void VoiceEnginesModel::setEngineParams( QModelIndex idx, int volume, int rate )
+{
+  if ( idx.isValid() )
+  {
+    voiceEngines[ idx.row() ].volume = volume;
+    voiceEngines[ idx.row() ].rate = rate;
+  }
+}
+
 VoiceEngineEditor::VoiceEngineEditor( SpeechClient::Engines const & engines, QWidget * parent ):
   QComboBox( parent )
 {

texttospeechsource.hh

@@ -29,10 +29,12 @@ public:
   VoiceEnginesModel( QWidget * parent, Config::VoiceEngines const & voiceEngines );
 
   void removeVoiceEngine( int index );
-  void addNewVoiceEngine( QString const & id, QString const & name );
+  void addNewVoiceEngine( QString const & id, QString const & name,
+                          int volume, int rate );
 
   Config::VoiceEngines const & getCurrentVoiceEngines() const
   { return voiceEngines; }
+  void setEngineParams( QModelIndex idx, int volume, int rate );
 
   QModelIndex index( int row, int column, QModelIndex const & parent ) const;
   QModelIndex parent( QModelIndex const & parent ) const;
@@ -94,12 +96,15 @@ private slots:
   void on_removeVoiceEngine_clicked();
   void on_previewVoice_clicked();
   void previewVoiceFinished();
+  void slidersChanged();
+  void selectionChanged();
 
 private:
   Ui::TextToSpeechSource ui;
   VoiceEnginesModel voiceEnginesModel;
 
   void fitSelectedVoiceEnginesColumns();
+  void adjustSliders();
 };
 
 #endif // __TEXTTOSPEECHSOURCE_HH_INCLUDED__

texttospeechsource.ui

@@ -78,6 +78,73 @@
      </item>
     </layout>
    </item>
+   <item>
+    <widget class="QGroupBox" name="groupBox">
+     <property name="title">
+      <string>Preferences</string>
+     </property>
+     <layout class="QHBoxLayout" name="horizontalLayout_5">
+      <item>
+       <widget class="QLabel" name="label_4">
+        <property name="text">
+         <string>Volume:</string>
+        </property>
+       </widget>
+      </item>
+      <item>
+       <widget class="QSlider" name="volumeSlider">
+        <property name="maximum">
+         <number>100</number>
+        </property>
+        <property name="sliderPosition">
+         <number>50</number>
+        </property>
+        <property name="tracking">
+         <bool>false</bool>
+        </property>
+        <property name="orientation">
+         <enum>Qt::Horizontal</enum>
+        </property>
+        <property name="tickPosition">
+         <enum>QSlider::TicksAbove</enum>
+        </property>
+        <property name="tickInterval">
+         <number>10</number>
+        </property>
+       </widget>
+      </item>
+      <item>
+       <widget class="QLabel" name="label_5">
+        <property name="text">
+         <string>Rate:</string>
+        </property>
+       </widget>
+      </item>
+      <item>
+       <widget class="QSlider" name="rateSlider">
+        <property name="maximum">
+         <number>100</number>
+        </property>
+        <property name="sliderPosition">
+         <number>50</number>
+        </property>
+        <property name="tracking">
+         <bool>false</bool>
+        </property>
+        <property name="orientation">
+         <enum>Qt::Horizontal</enum>
+        </property>
+        <property name="tickPosition">
+         <enum>QSlider::TicksAbove</enum>
+        </property>
+        <property name="tickInterval">
+         <number>10</number>
+        </property>
+       </widget>
+      </item>
+     </layout>
+    </widget>
+   </item>
    <item>
     <widget class="QGroupBox" name="groupBox_2">
      <property name="title">