Merge pull request #1256 from yqzhishen/add-dsdict-check

Add checks for phoneme definitions in dsdict.yaml

Author: stakira

OpenUtau.Core/DiffSinger/DiffSingerBasePhonemizer.cs

@@ -201,12 +201,19 @@ protected bool IsSyllableVowelExtensionNote(Note note) {
         /// distribute phonemes to each note inside the group
         /// </summary>
         List<phonemesPerNote> ProcessWord(Note[] notes, string[] symbols){
+            //Check if all phonemes are defined in dsdict.yaml (for their types)
+            foreach (var symbol in symbols) {
+                if (!g2p.IsValidSymbol(symbol)) {
+                    throw new InvalidDataException(
+                        $"Type definition of symbol \"{symbol}\" not found. Consider adding it to dsdict.yaml (or dsdict-<lang>.yaml) of the phonemizer.");
+                }
+            }
             var wordPhonemes = new List<phonemesPerNote>{
                 new phonemesPerNote(-1, notes[0].tone)
             };
             var dsPhonemes = symbols
                 .Select((symbol, index) => new dsPhoneme(symbol, GetSpeakerAtIndex(notes[0], index)))
-                .ToArray(); 
+                .ToArray();
             var isVowel = dsPhonemes.Select(s => g2p.IsVowel(s.Symbol)).ToArray();
             var isGlide = dsPhonemes.Select(s => g2p.IsGlide(s.Symbol)).ToArray();
             var nonExtensionNotes = notes.Where(n=>!IsSyllableVowelExtensionNote(n)).ToArray();

OpenUtau.Core/DiffSinger/DiffSingerPitch.cs

@@ -106,6 +106,15 @@ public RenderPitchResult Process(RenderPhrase phrase){
             var endMs = phrase.notes[^1].endMs + tailMs;
             int headFrames = (int)Math.Round(headMs / frameMs);
             int tailFrames = (int)Math.Round(tailMs / frameMs);
+            if (dsConfig.predict_dur || dsConfig.use_note_rest) {
+                //Check if all phonemes are defined in dsdict.yaml (for their types)
+                foreach (var phone in phrase.phones) {
+                    if (!g2p.IsValidSymbol(phone.phoneme)) {
+                        throw new InvalidDataException(
+                            $"Type definition of symbol \"{phone.phoneme}\" not found. Consider adding it to dsdict.yaml of the pitch predictor.");
+                    }
+                }
+            }
             //Linguistic Encoder
             var linguisticInputs = new List<NamedOnnxValue>();
             var tokens = phrase.phones

OpenUtau.Core/DiffSinger/DiffSingerVariance.cs

@@ -103,6 +103,15 @@ int PhonemeTokenize(string phoneme){
         public VarianceResult Process(RenderPhrase phrase){
             int headFrames = (int)Math.Round(headMs / frameMs);
             int tailFrames = (int)Math.Round(tailMs / frameMs);
+            if (dsConfig.predict_dur) {
+                //Check if all phonemes are defined in dsdict.yaml (for their types)
+                foreach (var phone in phrase.phones) {
+                    if (!g2p.IsValidSymbol(phone.phoneme)) {
+                        throw new InvalidDataException(
+                            $"Type definition of symbol \"{phone.phoneme}\" not found. Consider adding it to dsdict.yaml of the variance predictor.");
+                    }
+                }
+            }
             //Linguistic Encoder
             var linguisticInputs = new List<NamedOnnxValue>();
             var tokens = phrase.phones.Select(p => p.phoneme)