Merge pull request #22 from ShoukanLabs/dev

allow using legacy phonemisation where espeak dialect is needed

VoPho/engine.py

@@ -11,17 +11,22 @@ class Phonemizer:
     A class for phonemizing text in multiple languages,
     """
 
-    def __init__(self, working_path=None, stress=False):
+    def __init__(self, working_path=None, stress=False, legacy=False, manual_fixes=None):
         """
         Initialize the Phonemizer.
 
         :param working_path: Optional path for working directory
         :param stress: Optional toggle for stress, for phonemisers that support it
         """
+        if manual_fixes is None:
+            self.manual_fixes = {}
+        else:
+            self.manual_fixes = manual_fixes
         self.working_path = working_path
         self.stress = stress
         self._phonemizers = {}
         self.Tokenizer = Tokenizer()
+        self.legacy = legacy
 
     def pretty_print(self, tokens: list[Token]):
         """
@@ -61,13 +66,13 @@ def get_phonemizer(self, lang):
         """
         if lang not in self._phonemizers:
             if lang == 'en':
-                self._phonemizers[lang] = english.Phonemizer(stress=self.stress)
+                self._phonemizers[lang] = english.Phonemizer(stress=self.stress, legacy=self.legacy)
             elif lang == 'ja':
                 self._phonemizers[lang] = japanese.Phonemizer()
             elif lang == 'zh':
                 self._phonemizers[lang] = mandarin.Phonemizer()
             elif lang == 'cy': # cyrillic treated as russian
-                self._phonemizers[lang] = russian.Phonemizer(working_path=self.working_path)
+                self._phonemizers[lang] = russian.Phonemizer(working_path=self.working_path, stress=self.stress)
             elif lang == 'th':
                 self._phonemizers[lang] = thai.Phonemizer()
         return self._phonemizers.get(lang)

VoPho/phonemizers/english.py

@@ -291,12 +291,16 @@ def __call__(self, token):
 
 ### BASE PHONEMEISER CLASS
 class Phonemizer:
-    def __init__(self, manual_fixes=None, allow_heteronyms=True, stress=False):
+    def __init__(self, manual_fixes=None, allow_heteronyms=True, stress=False, legacy=False):
+        self.legacy = legacy
         if manual_fixes is None:
             manual_fixes = manual_phonemizations
-        self.backend = OpenPhonemizer()
-        self.fallback = OpenPhonemiserFallback(backend=self.backend)
-        self.phonemizer = en.G2P(trf=True, british=False, fallback=self.fallback)
+        if not legacy:
+            self.backend = OpenPhonemizer()
+            self.fallback = OpenPhonemiserFallback(backend=self.backend)
+            self.phonemizer = en.G2P(trf=True, british=False, fallback=self.fallback)
+        else:
+            self.phonemizer = OpenPhonemizer()
         self.manual_phonemizations = manual_fixes
         self.allow_heteronyms = allow_heteronyms
         self.stress = stress
@@ -338,9 +342,14 @@ def phonemize(self, text):
         segments = self.phoneme_tag_pattern.split(preprocessed_text)
 
         # Process each text segment (even indices) using the G2P model
-        for i in range(0, len(segments), 2):
-            if segments[i]:
-                segments[i] = self.phonemizer(segments[i])[0]
+        if not self.legacy:
+            for i in range(0, len(segments), 2):
+                if segments[i]:
+                    segments[i] = self.phonemizer(segments[i])[0]
+        else:
+            for i in range(0, len(segments), 2):
+                if segments[i]:
+                    segments[i] = self.phonemizer(segments[i])
 
         phonemized_text = ''.join(segments)
 
@@ -352,7 +361,7 @@ def phonemize(self, text):
 
 
 if __name__ == "__main__":
-    phonem = Phonemizer(stress=True)
+    phonem = Phonemizer(stress=True, legacy=True)
     test_text = "'two heads is better than one.', "
     print(f"Original: {test_text}")
     print(f"Phonemized: {phonem.phonemize(test_text)}")

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "VoPho"
-version = "0.0.18"
+version = "0.0.19"
 description = "An easy to use Multilingual phonemization meta-library"
 readme = "README.md"
 authors = [