phoonnx 0.2.6a2__py3-none-any.whl → 0.3.0__py3-none-any.whl

phoonnx/config.py

@@ -56,6 +56,7 @@ class PhonemeType(str, Enum):
     OPENPHONEMIZER = "openphonemizer" # en
     G2PEN = "g2pen" # en
 
+    TUGAPHONE = "tugaphone"  # pt
     G2PFA = "g2pfa"
     OPENJTALK = "openjtalk" # ja
     CUTLET = "cutlet" # ja
@@ -395,7 +396,7 @@ def get_phonemizer(phoneme_type: PhonemeType,
     from phoonnx.phonemizers import (EpitranPhonemizer, EspeakPhonemizer, OpenPhonemizer, OpenJTaklPhonemizer,
                                      ByT5Phonemizer, CharsiuPhonemizer, DeepPhonemizer, PersianPhonemizer,
                                      G2pCPhonemizer, G2pMPhonemizer, G2PKPhonemizer, G2PEnPhonemizer,
-                                     TransphonePhonemizer, MirandesePhonemizer, GoruutPhonemizer,
+                                     TransphonePhonemizer, MirandesePhonemizer, GoruutPhonemizer, TugaphonePhonemizer,
                                      GruutPhonemizer, GraphemePhonemizer, MantoqPhonemizer, MisakiPhonemizer,
                                      KoG2PPhonemizer, PypinyinPhonemizer, PyKakasiPhonemizer, CotoviaPhonemizer,
                                      CutletPhonemizer, PhonikudPhonemizer, VIPhonemePhonemizer, XpinyinPhonemizer,
@@ -404,6 +405,8 @@ def get_phonemizer(phoneme_type: PhonemeType,
         phonemizer = EspeakPhonemizer()
     elif phoneme_type == PhonemeType.BYT5:
         phonemizer = ByT5Phonemizer(model)
+    elif phoneme_type == PhonemeType.TUGAPHONE:
+        phonemizer = TugaphonePhonemizer()
     elif phoneme_type == PhonemeType.CHARSIU:
         phonemizer = CharsiuPhonemizer(model)
     elif phoneme_type == PhonemeType.GRUUT:

phoonnx/phonemizers/__init__.py

@@ -7,6 +7,7 @@ from phoonnx.phonemizers.vi import VIPhonemePhonemizer
 from phoonnx.phonemizers.he import PhonikudPhonemizer
 from phoonnx.phonemizers.ar import MantoqPhonemizer
 from phoonnx.phonemizers.fa import PersianPhonemizer
+from phoonnx.phonemizers.pt import TugaphonePhonemizer
 from phoonnx.phonemizers.ja import PyKakasiPhonemizer, CutletPhonemizer, OpenJTaklPhonemizer
 from phoonnx.phonemizers.ko import KoG2PPhonemizer, G2PKPhonemizer
 from phoonnx.phonemizers.zh import (G2pCPhonemizer, G2pMPhonemizer, PypinyinPhonemizer,
@@ -18,6 +19,7 @@ from phoonnx.phonemizers.mwl import MirandesePhonemizer
 Phonemizer = Union[
     MisakiPhonemizer,
     ByT5Phonemizer,
+    TugaphonePhonemizer,
     UnicodeCodepointPhonemizer,
     CharsiuPhonemizer,
     EspeakPhonemizer,

phoonnx/phonemizers/pt.py

@@ -0,0 +1,57 @@
+from phoonnx.phonemizers.base import BasePhonemizer
+from phoonnx.config import Alphabet
+
+
+class TugaphonePhonemizer(BasePhonemizer):
+
+    def __init__(self):
+        from tugaphone import TugaPhonemizer
+        self.tuga = TugaPhonemizer()
+        super().__init__(Alphabet.IPA)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["pt-PT", "pt-BR", "pt-AO", "pt-MZ", "pt-TL"])
+
+    def phonemize_string(self, text: str, lang: str) -> str:
+        lang = self.get_lang(lang)
+        # TODO - support regional dialects
+        return self.tuga.phonemize_sentence(text, lang)
+
+
+
+if __name__ == "__main__":
+
+    pho = TugaphonePhonemizer()
+
+    sentences = [
+        "O gato dorme.",
+        "Tu falas português muito bem.",
+        "O comboio chegou à estação.",
+        "A menina comeu o pão todo.",
+        "Vou pôr a manteiga no frigorífico.",
+        "Ele está a trabalhar no escritório.",
+        "Choveu muito ontem à noite.",
+        "A rapariga comprou um telemóvel novo.",
+        "Vamos tomar um pequeno-almoço.",
+        "O carro ficou sem gasolina."
+    ]
+
+    for s in sentences:
+        print(s)
+        for code in ["pt-PT", "pt-BR", "pt-AO", "pt-MZ", "pt-TL"]:
+            print(f"{code} → {pho.phonemize_string(s, code)}")
+        print("######")

phoonnx/version.py

@@ -1,8 +1,8 @@
 # START_VERSION_BLOCK
 VERSION_MAJOR = 0
-VERSION_MINOR = 2
-VERSION_BUILD = 6
-VERSION_ALPHA = 2
+VERSION_MINOR = 3
+VERSION_BUILD = 0
+VERSION_ALPHA = 0
 # END_VERSION_BLOCK
 
 VERSION_STR = f"{VERSION_MAJOR}.{VERSION_MINOR}.{VERSION_BUILD}"

{phoonnx-0.2.6a2.dist-info → phoonnx-0.3.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: phoonnx
-Version: 0.2.6a2
+Version: 0.3.0
 Home-page: https://github.com/TigreGotico/phoonnx
 Author: JarbasAi
 Author-email: jarbasai@mailfence.com
@@ -171,6 +171,7 @@ Requires-Dist: epitran; extra == "pl"
 Provides-Extra: pt
 Requires-Dist: gruut[pt]<3.0,>=2.3.0; extra == "pt"
 Requires-Dist: epitran; extra == "pt"
+Requires-Dist: tugaphone; extra == "pt"
 Provides-Extra: quy
 Requires-Dist: epitran; extra == "quy"
 Provides-Extra: rn

{phoonnx-0.2.6a2.dist-info → phoonnx-0.3.0.dist-info}/RECORD

@@ -1,14 +1,14 @@
 phoonnx/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-phoonnx/config.py,sha256=DKgsU03g8jrAuMcVqbu-w3MWPXOUihFtRnavg6WGQ1Y,19983
+phoonnx/config.py,sha256=WUhgICfj3JRv3KqTtFg3xrXM-ks_jY4Rx7Kyg7Lg5MI,20129
 phoonnx/phoneme_ids.py,sha256=FiNgZwV6naEsBh6XwFLh3_FyOgPiCsK9qo7S0v-CmI4,13667
 phoonnx/util.py,sha256=XSjFEoqSFcujFTHxednacgC9GrSYyF-Il5L6Utmxmu4,25909
-phoonnx/version.py,sha256=XtQ8LLEKPJw1x3fELi5iAV6qNgCNVf_oTmwVBTXYBqA,237
+phoonnx/version.py,sha256=NUwzOY8IGh0U8GKeNkPENyUypeBLWtHGxtzjSYT-4eE,237
 phoonnx/voice.py,sha256=Fh60o87-_sIw4zTP8NEGvEV7XzTC0FhW9Jkq7Fr60lQ,19382
 phoonnx/locale/ca/phonetic_spellings.txt,sha256=igv3t7jxLSRE5GHsdn57HOpxiWNcEmECPql6m02wbO0,47
 phoonnx/locale/en/phonetic_spellings.txt,sha256=xGQlWOABLzbttpQvopl9CU-NnwEJRqKx8iuylsdUoQA,27
 phoonnx/locale/gl/phonetic_spellings.txt,sha256=igv3t7jxLSRE5GHsdn57HOpxiWNcEmECPql6m02wbO0,47
 phoonnx/locale/pt/phonetic_spellings.txt,sha256=KntS8QMynEJ5A3Clvcjq4qlmL-ThSbhfD6v0nKSrlqs,49
-phoonnx/phonemizers/__init__.py,sha256=Ryregys3nWJrXJrOEJA-6xzOiXZWOCIwtYgd4KLW6UE,1816
+phoonnx/phonemizers/__init__.py,sha256=Na6ZjPJjPbf_MpQm0K7alxO-TlGBOIv9DOaal_m-xUo,1896
 phoonnx/phonemizers/ar.py,sha256=xxILq5iyH0kcI-NqFfRK4abGtpdUbykBjt_dZmPuO2w,3216
 phoonnx/phonemizers/base.py,sha256=FHvAsvSjAl_oSa1GoeEi96CQ_JO_xkKXWq0ukuMxiuo,8660
 phoonnx/phonemizers/en.py,sha256=N2SVoVhplQao7Ej5TXbxJU-YkAgkY0Fr9iYBFnsjFSE,9271
@@ -19,6 +19,7 @@ phoonnx/phonemizers/ja.py,sha256=Xojsrt715ihnIiEk9K6giYqDo9Iykw-SHfIidrHtHSU,383
 phoonnx/phonemizers/ko.py,sha256=kwWoOFqanCB8kv2JRx17A0hP78P1wbXlX6e8VBn1ezQ,2989
 phoonnx/phonemizers/mul.py,sha256=Y_M5BUY4Yka6Ba62Eea1HvgC6FTrrigaulo4KNRi1vE,99580
 phoonnx/phonemizers/mwl.py,sha256=xAOB1Bz_uVO14WbYlSFgvPxsezxzUKFwy6GT2mDgP2w,1007
+phoonnx/phonemizers/pt.py,sha256=6YCirAj5vTjf6SwpeMumCGVA9InexNMP1JZmG9P-TGc,1731
 phoonnx/phonemizers/vi.py,sha256=_XJc-Xeawr1Lxr7o8mE_hJao1aGcj4g01XYAOxC_Scg,1311
 phoonnx/phonemizers/zh.py,sha256=88Ywq8h9LDanlyz8RHjRSCY_PRK_Dq808tBADyrgaP8,9657
 phoonnx/thirdparty/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -63,7 +64,7 @@ phoonnx/thirdparty/tashkeel/input_id_map.json,sha256=cnpJqjx-k53AbzKyfC4GxMS771l
 phoonnx/thirdparty/tashkeel/model.onnx,sha256=UsQNQsoJT_n_B6CR0KHq_XuqXPI4jmCpzIm6zY5elV8,4788213
 phoonnx/thirdparty/tashkeel/target_id_map.json,sha256=baNAJL_UwP9U91mLt01aAEBRRNdGr-csFB_O6roh7TA,181
 phoonnx_train/export_onnx.py,sha256=UFn1id5-p9tIXxK9FvHOry59WkotNqXtupinkeKcSDs,12881
-phoonnx_train/preprocess.py,sha256=VYyU8XaCroa6tnLhQkQJ8wq1w1Gf9PPnIIM1N0PIcNE,22042
+phoonnx_train/preprocess.py,sha256=dAuHEF4-yEVhVpsVTRNat_S6_jljlwa4VpM5SZezyWc,23146
 phoonnx_train/train.py,sha256=nsINvDQ3dYvBne5UWPgLZ0a4qZFdSsOKk8HzZHGTLY4,8757
 phoonnx_train/norm_audio/__init__.py,sha256=Al_YwqMnENXRWp0c79cDZqbdd7pFYARXKxCfBaedr1c,3030
 phoonnx_train/norm_audio/trim.py,sha256=_ZsE3SYhahQSdEdBLeSwyFJGcvEbt-5E_lnWwTT4tcY,1698
@@ -83,7 +84,7 @@ phoonnx_train/vits/utils.py,sha256=exiyrtPHbnnGvcHWSbaH9-gR6srH5ZPHlKiqV2IHUrQ,4
 phoonnx_train/vits/wavfile.py,sha256=oQZiTIrdw0oLTbcVwKfGXye1WtKte6qK_52qVwiMvfc,26396
 phoonnx_train/vits/monotonic_align/__init__.py,sha256=5IdAOD1Z7UloMb6d_9NRFsXoNIjEQ3h9mvOSh_AtO3k,636
 phoonnx_train/vits/monotonic_align/setup.py,sha256=0K5iJJ2mKIklx6ncEfCQS34skm5hHPiz9vRlQEvevvY,266
-phoonnx-0.2.6a2.dist-info/METADATA,sha256=E3QnJVGElqhpgVBWDPL5ZT1PGH2fxB4lQaeXZYO99Kg,8250
-phoonnx-0.2.6a2.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
-phoonnx-0.2.6a2.dist-info/top_level.txt,sha256=ZrnHXe-4HqbOSX6fbdY-JiP7YEu2Bok9T0ji351MrmM,22
-phoonnx-0.2.6a2.dist-info/RECORD,,
+phoonnx-0.3.0.dist-info/METADATA,sha256=q8aeLB2Qlv5hLtryA399_EhE4pozB9lQMbVvQdmLQXk,8288
+phoonnx-0.3.0.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
+phoonnx-0.3.0.dist-info/top_level.txt,sha256=ZrnHXe-4HqbOSX6fbdY-JiP7YEu2Bok9T0ji351MrmM,22
+phoonnx-0.3.0.dist-info/RECORD,,

phoonnx_train/preprocess.py

@@ -377,6 +377,16 @@ def phonemize_worker(
     is_flag=True,
     help="Add diacritics to text (phonemizer specific, e.g., to denote stress).",
 )
+@click.option(
+    "--jsonl-audio-path",
+    default=None,
+    help="override audio_path base directory (everything before '/wav') in generated dataset.jsonl"
+)
+@click.option(
+    "--jsonl-audio-spec-path",
+    default=None,
+    help="override audio_norm_path/audio_spec_path base directory (everything before '/cache') in generated dataset.jsonl"
+)
 def cli(
     input_dir: Path,
     output_dir: Path,
@@ -397,6 +407,8 @@ def cli(
     skip_audio: bool,
     debug: bool,
     add_diacritics: bool,
+    jsonl_audio_path: Optional[str],
+    jsonl_audio_spec_path: Optional[str],
 ) -> None:
     """
     Preprocess a TTS dataset (e.g., LJSpeech format) for training a VITS-style model.
@@ -609,6 +621,17 @@ def cli(
                 _LOGGER.warning("Skipping utterance with invalid phoneme_ids before writing: %s", utt.audio_path)
                 continue
 
+            # apply path overrides if needed
+            # this allows pre-processing the dataset in one system and then train in other
+            if jsonl_audio_path:
+                base_path, fname = str(utt.audio_path).split("/wav/")
+                utt.audio_path = Path(f"{jsonl_audio_path}/wav/{fname}")
+            if jsonl_audio_spec_path:
+                base_path, fname = str(utt.audio_norm_path).split("/cache/")
+                utt.audio_norm_path = Path(f"{jsonl_audio_spec_path}/cache/{fname}")
+                base_path, fname = str(utt.audio_spec_path).split("/cache/")
+                utt.audio_spec_path = Path(f"{jsonl_audio_spec_path}/cache/{fname}")
+
             json.dump(
                 utt.asdict(),
                 dataset_file,