Change how end padding is applied, model JSON now has field to specif…

…y it
ZDisket · Nov 30, 2020 · 1914d5c · 1914d5c
1 parent 2459b8b
commit 1914d5c
Show file tree

Hide file tree

Showing 5 changed files with 9 additions and 12 deletions.
diff --git a/Voice.cpp b/Voice.cpp
@@ -135,7 +135,7 @@ void Voice::AddPhonemizer(Phonemizer *InPhn)
 std::vector<float> Voice::Vocalize(const std::string & Prompt, float Speed, int32_t SpeakerID, float Energy, float F0, int32_t EmotionID)
 {
 
-    std::string PhoneticTxt = Processor.ProcessTextPhonetic(Prompt,Phonemes,CurrentDict,(ETTSLanguage::Enum)VoxInfo.Language);
+    std::string PhoneticTxt = Processor.ProcessTextPhonetic(Prompt + VoxInfo.EndPadding,Phonemes,CurrentDict,(ETTSLanguage::Enum)VoxInfo.Language);
 
     TFTensor<float> Mel = MelPredictor.DoInference(PhonemesToID(PhoneticTxt), SpeakerID, Speed, Energy, F0,EmotionID);
 

diff --git a/VoxCommon.cpp b/VoxCommon.cpp
@@ -66,7 +66,8 @@ VoiceInfo VoxUtil::ReadModelJSON(const std::string &InfoFilename)
                  JS["note"].get<std::string>(),
                  JS["sarate"].get<uint32_t>(),
                  Lang,
-                LanguageNames[Lang]};
+                LanguageNames[Lang],
+                 " " + JS["pad"].get<std::string>()}; // Add a space for separation since we directly append the value to the prompt
 
     if (Inf.Note.size() > MaxNoteSize)
         Inf.Note = Inf.Note.substr(0,MaxNoteSize);

diff --git a/VoxCommon.hpp b/VoxCommon.hpp
@@ -79,6 +79,8 @@ struct VoiceInfo{
   uint32_t Language;
   std::string s_Language;
 
+  std::string EndPadding;
+
 
 
 };

diff --git a/mainwindow.cpp b/mainwindow.cpp
@@ -179,8 +179,6 @@ void MainWindow::on_btnInfer_clicked()
         on_btnLoad_clicked();
 
 
-
-
     // Convert to lowercase here before we add phonemes
     QString BeforeInput = ui->edtInput->toPlainText();
     QString RawInput = BeforeInput;
@@ -189,11 +187,6 @@ void MainWindow::on_btnInfer_clicked()
 
 
 
-
-
-
-
-
     QStringList InputSplits;
     QStringList BeforeSplits;
 
@@ -251,7 +244,7 @@ void MainWindow::on_btnInfer_clicked()
         Dets.Speed = RangeToFloat(ui->sliSpeed->value());
         Dets.Energy = RangeToFloat(ui->sliEnergy->value());
         Dets.pItem = widItm;
-        Dets.Prompt = idvInput + " @SIL @END";
+        Dets.Prompt = idvInput;
         Dets.SpeakerID = 0;
         Dets.EmotionID = -1;
         Dets.Denoise = ui->chkDenoise->isChecked();
@@ -322,6 +315,7 @@ void MainWindow::PlayBuffer(QBuffer *pBuff,bool ByUser)
 
     pBuff->open(QBuffer::ReadWrite);
 
+
     StdOutput->start(pBuff);
     CanPlayAudio = false;
 

diff --git a/phoneticdict.cpp b/phoneticdict.cpp
@@ -37,8 +37,8 @@ bool PhoneticDict::Import(const QString &infn)
 
 
     if (fi.GetFileLength() == 0){
-    fi.Close();
-    return true;
+        fi.Close();
+        return true;
 
     }
-Original file line number
+Diff line change
@@ Expand Up / @@ -79,6 +79,8 @@ struct VoiceInfo{ @@
       uint32_t Language;
       std::string s_Language;
+      std::string EndPadding;
     };
@@ Expand Down @@