Merge branch 'coqui-ai:main' into main

TTS/.models.json

@@ -650,6 +650,26 @@
                     "license": "CC-BY-4.0"
                 }
             }
+        },
+        "bn":{
+            "custom":{
+                "vits-male":{
+                    "github_rls_url":"https://coqui.gateway.scarf.sh/v0.13.3_models/tts_models--bn--custom--vits_male.zip",
+                    "default_vocoder": null,
+                    "commit": null,
+                    "description": "Single speaker Bangla male model. For more information -> https://github.com/mobassir94/comprehensive-bangla-tts",
+                    "author": "@mobassir94",
+                    "license": "Apache 2.0"
+                },
+                "vits-female":{
+                    "github_rls_url":"https://coqui.gateway.scarf.sh/v0.13.3_models/tts_models--bn--custom--vits_female.zip",
+                    "default_vocoder": null,
+                    "commit": null,
+                    "description": "Single speaker Bangla female model. For more information -> https://github.com/mobassir94/comprehensive-bangla-tts",
+                    "author": "@mobassir94",
+                    "license": "Apache 2.0"
+                }
+            }
         }
     },
     "vocoder_models": {

TTS/VERSION

@@ -1 +1 @@
-0.13.2
+0.13.3

TTS/api.py

@@ -303,6 +303,8 @@ class TTS:
 
     @property
     def is_coqui_studio(self):
+        if self.model_name is None:
+            return False
         return "coqui_studio" in self.model_name
 
     @property

TTS/tts/utils/text/bangla/phonemizer.py

@@ -0,0 +1,121 @@
+import re
+
+import bangla
+from bnnumerizer import numerize
+from bnunicodenormalizer import Normalizer
+
+# initialize
+bnorm = Normalizer()
+
+
+attribution_dict = {
+    "সাঃ": "সাল্লাল্লাহু আলাইহি ওয়া সাল্লাম",
+    "আঃ": "আলাইহিস সালাম",
+    "রাঃ": "রাদিআল্লাহু আনহু",
+    "রহঃ": "রহমাতুল্লাহি আলাইহি",
+    "রহিঃ": "রহিমাহুল্লাহ",
+    "হাফিঃ": "হাফিযাহুল্লাহ",
+    "বায়ান": "বাইআন",
+    "দাঃবাঃ": "দামাত বারাকাতুহুম,দামাত বারাকাতুল্লাহ",
+    # "আয়াত" : "আইআত",#আইআত
+    # "ওয়া" : "ওআ",
+    # "ওয়াসাল্লাম"  : "ওআসাল্লাম",
+    # "কেন"  : "কেনো",
+    # "কোন" : "কোনো",
+    # "বল"   : "বলো",
+    # "চল"   : "চলো",
+    # "কর"   : "করো",
+    # "রাখ"   : "রাখো",
+    "’": "",
+    "‘": "",
+    # "য়"     : "অ",
+    # "সম্প্রদায়" : "সম্প্রদাই",
+    # "রয়েছে"   : "রইছে",
+    # "রয়েছ"    : "রইছ",
+    "/": " বাই ",
+}
+
+
+def tag_text(text: str):
+    # remove multiple spaces
+    text = re.sub(" +", " ", text)
+    # create start and end
+    text = "start" + text + "end"
+    # tag text
+    parts = re.split("[\u0600-\u06FF]+", text)
+    # remove non chars
+    parts = [p for p in parts if p.strip()]
+    # unique parts
+    parts = set(parts)
+    # tag the text
+    for m in parts:
+        if len(m.strip()) > 1:
+            text = text.replace(m, f"{m}")
+    # clean-tags
+    text = text.replace("start", "")
+    text = text.replace("end", "")
+    return text
+
+
+def normalize(sen):
+    global bnorm  # pylint: disable=global-statement
+    _words = [bnorm(word)["normalized"] for word in sen.split()]
+    return " ".join([word for word in _words if word is not None])
+
+
+def expand_full_attribution(text):
+    for word, attr in attribution_dict.items():
+        if word in text:
+            text = text.replace(word, normalize(attr))
+    return text
+
+
+def collapse_whitespace(text):
+    # Regular expression matching whitespace:
+    _whitespace_re = re.compile(r"\s+")
+    return re.sub(_whitespace_re, " ", text)
+
+
+def bangla_text_to_phonemes(text: str) -> str:
+    # english numbers to bangla conversion
+    res = re.search("[0-9]", text)
+    if res is not None:
+        text = bangla.convert_english_digit_to_bangla_digit(text)
+
+    # replace ':' in between two bangla numbers with ' এর '
+    pattern = r"[০, ১, ২, ৩, ৪, ৫, ৬, ৭, ৮, ৯]:[০, ১, ২, ৩, ৪, ৫, ৬, ৭, ৮, ৯]"
+    matches = re.findall(pattern, text)
+    for m in matches:
+        r = m.replace(":", " এর ")
+        text = text.replace(m, r)
+
+    # numerize text
+    text = numerize(text)
+
+    # tag sections
+    text = tag_text(text)
+
+    # text blocks
+    # blocks = text.split("")
+    # blocks = [b for b in blocks if b.strip()]
+
+    # create tuple of (lang,text)
+    if "" in text:
+        text = text.replace("", "").replace("", "")
+    # Split based on sentence ending Characters
+    bn_text = text.strip()
+
+    sentenceEnders = re.compile("[।!?]")
+    sentences = sentenceEnders.split(str(bn_text))
+
+    data = ""
+    for sent in sentences:
+        res = re.sub("\n", "", sent)
+        res = normalize(res)
+        # expand attributes
+        res = expand_full_attribution(res)
+
+        res = collapse_whitespace(res)
+        res += "।"
+        data += res
+    return data

TTS/tts/utils/text/phonemizers/__init__.py

@@ -1,3 +1,4 @@
+from TTS.tts.utils.text.phonemizers.bangla_phonemizer import BN_Phonemizer
 from TTS.tts.utils.text.phonemizers.base import BasePhonemizer
 from TTS.tts.utils.text.phonemizers.espeak_wrapper import ESpeak
 from TTS.tts.utils.text.phonemizers.gruut_wrapper import Gruut
@@ -28,6 +29,7 @@ DEF_LANG_TO_PHONEMIZER["en"] = DEF_LANG_TO_PHONEMIZER["en-us"]
 DEF_LANG_TO_PHONEMIZER["ja-jp"] = JA_JP_Phonemizer.name()
 DEF_LANG_TO_PHONEMIZER["zh-cn"] = ZH_CN_Phonemizer.name()
 DEF_LANG_TO_PHONEMIZER["ko-kr"] = KO_KR_Phonemizer.name()
+DEF_LANG_TO_PHONEMIZER["bn"] = BN_Phonemizer.name()
 
 
 def get_phonemizer_by_name(name: str, **kwargs) -> BasePhonemizer:
@@ -50,6 +52,8 @@ def get_phonemizer_by_name(name: str, **kwargs) -> BasePhonemizer:
         return JA_JP_Phonemizer(**kwargs)
     if name == "ko_kr_phonemizer":
         return KO_KR_Phonemizer(**kwargs)
+    if name == "bn_phonemizer":
+        return BN_Phonemizer(**kwargs)
     raise ValueError(f"Phonemizer {name} not found")
 
 

TTS/tts/utils/text/phonemizers/bangla_phonemizer.py

@@ -0,0 +1,62 @@
+from typing import Dict
+
+from TTS.tts.utils.text.bangla.phonemizer import bangla_text_to_phonemes
+from TTS.tts.utils.text.phonemizers.base import BasePhonemizer
+
+_DEF_ZH_PUNCS = "、.,[]()?!〽~『』「」【】"
+
+
+class BN_Phonemizer(BasePhonemizer):
+    """🐸TTS bn phonemizer using functions in `TTS.tts.utils.text.bangla.phonemizer`
+
+    Args:
+        punctuations (str):
+            Set of characters to be treated as punctuation. Defaults to `_DEF_ZH_PUNCS`.
+
+        keep_puncs (bool):
+            If True, keep the punctuations after phonemization. Defaults to False.
+
+    Example ::
+
+        "这是，样本中文。" -> `d|ʒ|ø|4| |ʂ|ʏ|4| |，| |i|ɑ|ŋ|4|b|œ|n|3| |d|ʒ|o|ŋ|1|w|œ|n|2| |。`
+
+    TODO: someone with Bangla knowledge should check this implementation
+    """
+
+    language = "bn"
+
+    def __init__(self, punctuations=_DEF_ZH_PUNCS, keep_puncs=False, **kwargs):  # pylint: disable=unused-argument
+        super().__init__(self.language, punctuations=punctuations, keep_puncs=keep_puncs)
+
+    @staticmethod
+    def name():
+        return "bn_phonemizer"
+
+    @staticmethod
+    def phonemize_bn(text: str, separator: str = "|") -> str:  # pylint: disable=unused-argument
+        ph = bangla_text_to_phonemes(text)
+        return ph
+
+    def _phonemize(self, text, separator):
+        return self.phonemize_bn(text, separator)
+
+    @staticmethod
+    def supported_languages() -> Dict:
+        return {"bn": "Bangla"}
+
+    def version(self) -> str:
+        return "0.0.1"
+
+    def is_available(self) -> bool:
+        return True
+
+
+if __name__ == "__main__":
+    txt = "রাসূলুল্লাহ সাল্লাল্লাহু আলাইহি ওয়া সাল্লাম শিক্ষা দিয়েছেন যে, কেউ যদি কোন খারাপ কিছুর সম্মুখীন হয়, তখনও যেন বলে."
+    e = BN_Phonemizer()
+    print(e.supported_languages())
+    print(e.version())
+    print(e.language)
+    print(e.name())
+    print(e.is_available())
+    print("`" + e.phonemize(txt) + "`")

docs/source/implementing_a_new_language_frontend.md

@@ -0,0 +1,12 @@
+# Implementing a New Language Frontend
+
+- Language frontends are located under `TTS.tts.utils.text`
+- Each special language has a separate folder.
+- Each folder containst all the utilities for processing the text input.
+- `TTS.tts.utils.text.phonemizers` contains the main phonemizer for a language. This is the class that uses the utilities
+from the previous step and used to convert the text to phonemes or graphemes for the model.
+- After you implement your phonemizer, you need to add it to the `TTS/tts/utils/text/phonemizers/__init__.py` to be able to
+map the language code in the model config - `config.phoneme_language` - to the phonemizer class and initiate the phonemizer automatically.
+- You should also add tests to `tests/text_tests` if you want to make a PR.
+
+We suggest you to check the available implementations as reference. Good luck!

docs/source/index.md

@@ -22,6 +22,7 @@
     inference
     docker_images
     implementing_a_new_model
+    implementing_a_new_language_frontend
     training_a_model
     finetuning
     configuration

requirements.txt

@@ -41,3 +41,7 @@ gruut[de]==2.2.3
 jamo
 nltk
 g2pkk>=0.1.1
+# deps for bangla
+bangla==0.0.2
+bnnumerizer
+bnunicodenormalizer==0.1.1

tests/text_tests/test_phonemizer.py

@@ -3,6 +3,7 @@ import unittest
 from packaging.version import Version
 
 from TTS.tts.utils.text.phonemizers import ESpeak, Gruut, JA_JP_Phonemizer, ZH_CN_Phonemizer
+from TTS.tts.utils.text.phonemizers.bangla_phonemizer import BN_Phonemizer
 from TTS.tts.utils.text.phonemizers.multi_phonemizer import MultiPhonemizer
 
 EXAMPLE_TEXTs = [
@@ -230,6 +231,28 @@ class TestZH_CN_Phonemizer(unittest.TestCase):
         self.assertTrue(self.phonemizer.is_available())
 
 
+class TestBN_Phonemizer(unittest.TestCase):
+    def setUp(self):
+        self.phonemizer = BN_Phonemizer()
+        self._TEST_CASES = "রাসূলুল্লাহ সাল্লাল্লাহু আলাইহি ওয়া সাল্লাম শিক্ষা দিয়েছেন যে, কেউ যদি কোন খারাপ কিছুর সম্মুখীন হয়, তখনও যেন"
+        self._EXPECTED = "রাসূলুল্লাহ সাল্লাল্লাহু আলাইহি ওয়া সাল্লাম শিক্ষা দিয়েছেন যে কেউ যদি কোন খারাপ কিছুর সম্মুখীন হয় তখনও যেন।"
+
+    def test_phonemize(self):
+        self.assertEqual(self.phonemizer.phonemize(self._TEST_CASES, separator=""), self._EXPECTED)
+
+    def test_name(self):
+        self.assertEqual(self.phonemizer.name(), "bn_phonemizer")
+
+    def test_get_supported_languages(self):
+        self.assertIsInstance(self.phonemizer.supported_languages(), dict)
+
+    def test_get_version(self):
+        self.assertIsInstance(self.phonemizer.version(), str)
+
+    def test_is_available(self):
+        self.assertTrue(self.phonemizer.is_available())
+
+
 class TestMultiPhonemizer(unittest.TestCase):
     def setUp(self):
         self.phonemizer = MultiPhonemizer({"tr": "espeak", "en-us": "", "de": "gruut", "zh-cn": ""})