Merge branch 'dev' of https://github.com/mozilla/TTS into dev

TTS/.models.json

@@ -75,6 +75,16 @@
                     "contact":"erengolge@gmail.com"
                 }
             }
+        },
+        "zh":{
+            "baker":{
+                "tacotron2-DDC-GST":{
+                    "model_file": "1RR9rZdV_FMm8yvtCHALtUbJf1nxbUiAw",
+                    "config_file": "1daY1JHGXEozJ-MGYLiWEUmzEwEvM5xpz",
+                    "stats_file": "1vl9c-D3dW_E7pdhNpDFQLX-giJc0jOtV",
+                    "commit": ""
+                }
+            }
         }
     },
     "vocoder_models":{

TTS/tts/datasets/preprocess.py

@@ -3,6 +3,7 @@ from glob import glob
 import re
 import sys
 from pathlib import Path
+from typing import List, Tuple
 
 from tqdm import tqdm
 
@@ -368,3 +369,24 @@ def _voxcel_x(root_path, meta_file, voxcel_idx):
 
     with open(str(cache_to), 'r') as f:
         return [x.strip().split('|') for x in f.readlines()]
+
+
+
+def baker(root_path: str, meta_file: str) ->  List[List[str]]:
+    """Normalizes the Baker meta data file to TTS format
+
+    Args:
+        root_path (str): path to the baker dataset
+        meta_file (str): name of the meta dataset containing names of wav to select and the transcript of the sentence  
+    Returns:
+        List[List[str]]: List of (text, wav_path, speaker_name) associated with each sentences
+    """
+    txt_file = os.path.join(root_path, meta_file)
+    items = []
+    speaker_name = "baker"
+    with open(txt_file, 'r') as ttf:
+        for line in ttf:
+            wav_name, text = line.rstrip('\n').split("|")
+            wav_path = os.path.join(root_path, "clips_22", wav_name)
+            items.append([text, wav_path, speaker_name])
+    return items 

TTS/tts/utils/chinese_mandarin/numbers.py

@@ -0,0 +1,131 @@
+
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+
+# Licensed under WTFPL or the Unlicense or CC0.
+# This uses Python 3, but it's easy to port to Python 2 by changing
+# strings to u'xx'.
+
+import re
+import itertools
+
+
+def _num2chinese(num :str, big=False, simp=True, o=False, twoalt=False) -> str:
+    """Convert numerical arabic numbers (0->9) to chinese hanzi numbers (〇 -> 九)
+
+    Args:
+        num (str): arabic number to convert
+        big (bool, optional): use financial characters. Defaults to False.
+        simp (bool, optional): use simplified characters instead of tradictional characters. Defaults to True.
+        o (bool, optional): use 〇 for 'zero'. Defaults to False.
+        twoalt (bool, optional): use 两/兩 for 'two' when appropriate. Defaults to False.
+
+    Raises:
+        ValueError: if number is more than 1e48
+        ValueError: if 'e' exposent in number
+
+    Returns:
+        str: converted number as hanzi characters
+    """
+
+    # check num first
+    nd = str(num)
+    if abs(float(nd)) >= 1e48:
+        raise ValueError('number out of range')
+    elif 'e' in nd:
+        raise ValueError('scientific notation is not supported')
+    c_symbol = '正负点' if simp else '正負點'
+    if o:  # formal
+        twoalt = False
+    if big:
+        c_basic = '零壹贰叁肆伍陆柒捌玖' if simp else '零壹貳參肆伍陸柒捌玖'
+        c_unit1 = '拾佰仟'
+        c_twoalt = '贰' if simp else '貳'
+    else:
+        c_basic = '〇一二三四五六七八九' if o else '零一二三四五六七八九'
+        c_unit1 = '十百千'
+        if twoalt:
+            c_twoalt = '两' if simp else '兩'
+        else:
+            c_twoalt = '二'
+    c_unit2 = '万亿兆京垓秭穰沟涧正载' if simp else '萬億兆京垓秭穰溝澗正載'
+    revuniq = lambda l: ''.join(k for k, g in itertools.groupby(reversed(l)))
+    nd = str(num)
+    result = []
+    if nd[0] == '+':
+        result.append(c_symbol[0])
+    elif nd[0] == '-':
+        result.append(c_symbol[1])
+    if '.' in nd:
+        integer, remainder = nd.lstrip('+-').split('.')
+    else:
+        integer, remainder = nd.lstrip('+-'), None
+    if int(integer):
+        splitted = [integer[max(i - 4, 0):i]
+                    for i in range(len(integer), 0, -4)]
+        intresult = []
+        for nu, unit in enumerate(splitted):
+            # special cases
+            if int(unit) == 0:  # 0000
+                intresult.append(c_basic[0])
+                continue
+            elif nu > 0 and int(unit) == 2:  # 0002
+                intresult.append(c_twoalt + c_unit2[nu - 1])
+                continue
+            ulist = []
+            unit = unit.zfill(4)
+            for nc, ch in enumerate(reversed(unit)):
+                if ch == '0':
+                    if ulist:  # ???0
+                        ulist.append(c_basic[0])
+                elif nc == 0:
+                    ulist.append(c_basic[int(ch)])
+                elif nc == 1 and ch == '1' and unit[1] == '0':
+                    # special case for tens
+                    # edit the 'elif' if you don't like
+                    # 十四, 三千零十四, 三千三百一十四
+                    ulist.append(c_unit1[0])
+                elif nc > 1 and ch == '2':
+                    ulist.append(c_twoalt + c_unit1[nc - 1])
+                else:
+                    ulist.append(c_basic[int(ch)] + c_unit1[nc - 1])
+            ustr = revuniq(ulist)
+            if nu == 0:
+                intresult.append(ustr)
+            else:
+                intresult.append(ustr + c_unit2[nu - 1])
+        result.append(revuniq(intresult).strip(c_basic[0]))
+    else:
+        result.append(c_basic[0])
+    if remainder:
+        result.append(c_symbol[2])
+        result.append(''.join(c_basic[int(ch)] for ch in remainder))
+    return ''.join(result)
+
+
+
+
+def _number_replace(match) -> str:
+    """function to apply in a match, transform all numbers in a match by chinese characters
+
+    Args:
+        match (re.Match): numbers regex matches
+
+    Returns:
+        str: replaced characters for the numbers
+    """
+    match_str: str = match.group()
+    return _num2chinese(match_str)
+
+
+def replace_numbers_to_characters_in_text(text: str) -> str:
+    """Replace all arabic numbers in a text by their equivalent in chinese characters (simplified)
+
+    Args:
+        text (str): input text to transform
+
+    Returns:
+        str: output text
+    """
+    text = re.sub(r'[0-9]+', _number_replace, text)
+    return text

TTS/tts/utils/chinese_mandarin/phonemizer.py

@@ -0,0 +1,41 @@
+from typing import List
+
+import pypinyin
+
+from .pinyinToPhonemes import PINYIN_DICT
+
+
+import jieba
+
+
+def _chinese_character_to_pinyin(text: str) -> List[str]:
+    pinyins = pypinyin.pinyin(
+        text, style=pypinyin.Style.TONE3, heteronym=False, neutral_tone_with_five=True
+    )
+    pinyins_flat_list = [item for sublist in pinyins for item in sublist]
+    return pinyins_flat_list
+
+
+def _chinese_pinyin_to_phoneme(pinyin: str) -> str:
+    segment = pinyin[:-1]
+    tone = pinyin[-1]
+    phoneme = PINYIN_DICT.get(segment, [""])[0]
+    return phoneme + tone
+
+
+def chinese_text_to_phonemes(text: str) -> str:
+    tokenized_text = jieba.cut(text, HMM=False)
+    tokenized_text = " ".join(tokenized_text)
+    pinyined_text: List[str] = _chinese_character_to_pinyin(tokenized_text)
+
+    results: List[str] = []
+
+    for token in pinyined_text:
+        if token[-1] in "12345":  # TODO transform to is_pinyin()
+            pinyin_phonemes = _chinese_pinyin_to_phoneme(token)
+
+            results += list(pinyin_phonemes)
+        else:  # is ponctuation or other
+            results += list(token)
+
+    return "|".join(results)

TTS/tts/utils/chinese_mandarin/pinyinToPhonemes.py

@@ -0,0 +1,420 @@
+
+PINYIN_DICT = {
+    "a": ["a"],
+    "ai": ["ai"],
+    "an": ["an"],
+    "ang": ["ɑŋ"],
+    "ao": ["aʌ"],
+    "ba": ["ba"],
+    "bai": ["bai"],
+    "ban": ["ban"],
+    "bang": ["bɑŋ"],
+    "bao": ["baʌ"], 
+    # "be": ["be"], doesnt exist
+    "bei": ["bɛi"],
+    "ben": ["bœn"],
+    "beng": ["bɵŋ"],
+    "bi": ["bi"],
+    "bian": ["biɛn"],
+    "biao": ["biaʌ"],
+    "bie": ["bie"],
+    "bin": ["bin"],
+    "bing": ["bɨŋ"],
+    "bo": ["bo"],
+    "bu": ["bu"],
+    "ca": ["tsa"],
+    "cai": ["tsai"],
+    "can": ["tsan"],
+    "cang": ["tsɑŋ"],
+    "cao": ["tsaʌ"],
+    "ce": ["tsø"],
+    "cen": ["tsœn"],
+    "ceng": ["tsɵŋ"],
+    "cha": ["ʈʂa"],
+    "chai": ["ʈʂai"],
+    "chan": ["ʈʂan"],
+    "chang": ["ʈʂɑŋ"],
+    "chao": ["ʈʂaʌ"],
+    "che": ["ʈʂø"],
+    "chen": ["ʈʂœn"],
+    "cheng": ["ʈʂɵŋ"],
+    "chi": ["ʈʂʏ"],
+    "chong": ["ʈʂoŋ"],
+    "chou": ["ʈʂou"],
+    "chu": ["ʈʂu"],
+    "chua": ["ʈʂua"],
+    "chuai": ["ʈʂuai"],
+    "chuan": ["ʈʂuan"],
+    "chuang": ["ʈʂuɑŋ"],
+    "chui": ["ʈʂuei"],
+    "chun": ["ʈʂun"],
+    "chuo": ["ʈʂuo"],
+    "ci": ["tsɪ"],
+    "cong": ["tsoŋ"],
+    "cou": ["tsou"],
+    "cu": ["tsu"],
+    "cuan": ["tsuan"],
+    "cui": ["tsuei"],
+    "cun": ["tsun"],
+    "cuo": ["tsuo"],
+    "da": ["da"],
+    "dai": ["dai"],
+    "dan": ["dan"],
+    "dang": ["dɑŋ"],
+    "dao": ["daʌ"],
+    "de": ["dø"],
+    "dei": ["dei"],
+    # "den": ["dœn"],
+    "deng": ["dɵŋ"],
+    "di": ["di"],
+    "dia": ["dia"],
+    "dian": ["diɛn"],
+    "diao": ["diaʌ"],
+    "die": ["die"],
+    "ding": ["dɨŋ"],
+    "diu": ["dio"],
+    "dong": ["doŋ"],
+    "dou": ["dou"],
+    "du": ["du"],
+    "duan": ["duan"],
+    "dui": ["duei"],
+    "dun": ["dun"],
+    "duo": ["duo"],
+    "e": ["ø"],
+    "ei": ["ei"],
+    "en": ["œn"],
+    # "ng": ["œn"],
+    # "eng": ["ɵŋ"],
+    "er": ["er"],
+    "fa": ["fa"],
+    "fan": ["fan"],
+    "fang": ["fɑŋ"],
+    "fei": ["fei"],
+    "fen": ["fœn"],
+    "feng": ["fɵŋ"],
+    "fo": ["fo"],
+    "fou": ["fou"],
+    "fu": ["fu"],
+    "ga": ["ga"],
+    "gai": ["gai"],
+    "gan": ["gan"],
+    "gang": ["gɑŋ"],
+    "gao": ["gaʌ"],
+    "ge": ["gø"],
+    "gei": ["gei"],
+    "gen": ["gœn"],
+    "geng": ["gɵŋ"],
+    "gong": ["goŋ"],
+    "gou": ["gou"],
+    "gu": ["gu"],
+    "gua": ["gua"],
+    "guai": ["guai"],
+    "guan": ["guan"],
+    "guang": ["guɑŋ"],
+    "gui": ["guei"],
+    "gun": ["gun"],
+    "guo": ["guo"],
+    "ha": ["xa"],
+    "hai": ["xai"],
+    "han": ["xan"],
+    "hang": ["xɑŋ"],
+    "hao": ["xaʌ"],
+    "he": ["xø"],
+    "hei": ["xei"],
+    "hen": ["xœn"],
+    "heng": ["xɵŋ"],
+    "hong": ["xoŋ"],
+    "hou": ["xou"],
+    "hu": ["xu"],
+    "hua": ["xua"],
+    "huai": ["xuai"],
+    "huan": ["xuan"],
+    "huang": ["xuɑŋ"],
+    "hui": ["xuei"],
+    "hun": ["xun"],
+    "huo": ["xuo"],
+    "ji": ["dʑi"],
+    "jia": ["dʑia"],
+    "jian": ["dʑiɛn"],
+    "jiang": ["dʑiɑŋ"],
+    "jiao": ["dʑiaʌ"],
+    "jie": ["dʑie"],
+    "jin": ["dʑin"],
+    "jing": ["dʑɨŋ"],
+    "jiong": ["dʑioŋ"],
+    "jiu": ["dʑio"],
+    "ju": ["dʑy"],
+    "juan": ["dʑyɛn"],
+    "jue": ["dʑye"],
+    "jun": ["dʑyn"],
+    "ka": ["ka"],
+    "kai": ["kai"],
+    "kan": ["kan"],
+    "kang": ["kɑŋ"],
+    "kao": ["kaʌ"],
+    "ke": ["kø"],
+    "kei": ["kei"],
+    "ken": ["kœn"],
+    "keng": ["kɵŋ"],
+    "kong": ["koŋ"],
+    "kou": ["kou"],
+    "ku": ["ku"],
+    "kua": ["kua"],
+    "kuai": ["kuai"],
+    "kuan": ["kuan"],
+    "kuang": ["kuɑŋ"],
+    "kui": ["kuei"],
+    "kun": ["kun"],
+    "kuo": ["kuo"],
+    "la": ["la"],
+    "lai": ["lai"],
+    "lan": ["lan"],
+    "lang": ["lɑŋ"],
+    "lao": ["laʌ"],
+    "le": ["lø"],
+    "lei": ["lei"],
+    "leng": ["lɵŋ"],
+    "li": ["li"],
+    "lia": ["lia"],
+    "lian": ["liɛn"],
+    "liang": ["liɑŋ"],
+    "liao": ["liaʌ"],
+    "lie": ["lie"],
+    "lin": ["lin"],
+    "ling": ["lɨŋ"],
+    "liu": ["lio"],
+    "lo": ["lo"],
+    "long": ["loŋ"],
+    "lou": ["lou"],
+    "lu": ["lu"],
+    "lv": ["ly"],
+    "luan": ["luan"],
+    "lve": ["lye"],
+    "lue": ["lue"],
+    "lun": ["lun"],
+    "luo": ["luo"],
+    "ma": ["ma"],
+    "mai": ["mai"],
+    "man": ["man"],
+    "mang": ["mɑŋ"],
+    "mao": ["maʌ"],
+    "me": ["mø"],
+    "mei": ["mei"],
+    "men": ["mœn"],
+    "meng": ["mɵŋ"],
+    "mi": ["mi"],
+    "mian": ["miɛn"],
+    "miao": ["miaʌ"],
+    "mie": ["mie"],
+    "min": ["min"],
+    "ming": ["mɨŋ"],
+    "miu": ["mio"],
+    "mo": ["mo"],
+    "mou": ["mou"],
+    "mu": ["mu"],
+    "na": ["na"],
+    "nai": ["nai"],
+    "nan": ["nan"],
+    "nang": ["nɑŋ"],
+    "nao": ["naʌ"],
+    "ne": ["nø"],
+    "nei": ["nei"],
+    "nen": ["nœn"],
+    "neng": ["nɵŋ"],
+    "ni": ["ni"],
+    "nia": ["nia"],
+    "nian": ["niɛn"],
+    "niang": ["niɑŋ"],
+    "niao": ["niaʌ"],
+    "nie": ["nie"],
+    "nin": ["nin"],
+    "ning": ["nɨŋ"],
+    "niu": ["nio"],
+    "nong": ["noŋ"],
+    "nou": ["nou"],
+    "nu": ["nu"],
+    "nv": ["ny"],
+    "nuan": ["nuan"],
+    "nve": ["nye"],
+    "nue": ["nye"],
+    "nuo": ["nuo"],
+    "o": ["o"],
+    "ou": ["ou"],
+    "pa": ["pa"],
+    "pai": ["pai"],
+    "pan": ["pan"],
+    "pang": ["pɑŋ"],
+    "pao": ["paʌ"],
+    "pe": ["pø"],
+    "pei": ["pei"],
+    "pen": ["pœn"],
+    "peng": ["pɵŋ"],
+    "pi": ["pi"],
+    "pian": ["piɛn"],
+    "piao": ["piaʌ"],
+    "pie": ["pie"],
+    "pin": ["pin"],
+    "ping": ["pɨŋ"],
+    "po": ["po"],
+    "pou": ["pou"],
+    "pu": ["pu"],
+    "qi": ["tɕi"],
+    "qia": ["tɕia"],
+    "qian": ["tɕiɛn"],
+    "qiang": ["tɕiɑŋ"],
+    "qiao": ["tɕiaʌ"],
+    "qie": ["tɕie"],
+    "qin": ["tɕin"],
+    "qing": ["tɕɨŋ"],
+    "qiong": ["tɕioŋ"],
+    "qiu": ["tɕio"],
+    "qu": ["tɕy"],
+    "quan": ["tɕyɛn"],
+    "que": ["tɕye"],
+    "qun": ["tɕyn"],
+    "ran": ["ʐan"],
+    "rang": ["ʐɑŋ"],
+    "rao": ["ʐaʌ"],
+    "re": ["ʐø"],
+    "ren": ["ʐœn"],
+    "reng": ["ʐɵŋ"],
+    "ri": ["ʐʏ"],
+    "rong": ["ʐoŋ"],
+    "rou": ["ʐou"],
+    "ru": ["ʐu"],
+    "rua": ["ʐua"],
+    "ruan": ["ʐuan"],
+    "rui": ["ʐuei"],
+    "run": ["ʐun"],
+    "ruo": ["ʐuo"],
+    "sa": ["sa"],
+    "sai": ["sai"],
+    "san": ["san"],
+    "sang": ["sɑŋ"],
+    "sao": ["saʌ"],
+    "se": ["sø"],
+    "sen": ["sœn"],
+    "seng": ["sɵŋ"],
+    "sha": ["ʂa"],
+    "shai": ["ʂai"],
+    "shan": ["ʂan"],
+    "shang": ["ʂɑŋ"],
+    "shao": ["ʂaʌ"],
+    "she": ["ʂø"],
+    "shei": ["ʂei"],
+    "shen": ["ʂœn"],
+    "sheng": ["ʂɵŋ"],
+    "shi": ["ʂʏ"],
+    "shou": ["ʂou"],
+    "shu": ["ʂu"],
+    "shua": ["ʂua"],
+    "shuai": ["ʂuai"],
+    "shuan": ["ʂuan"],
+    "shuang": ["ʂuɑŋ"],
+    "shui": ["ʂuei"],
+    "shun": ["ʂun"],
+    "shuo": ["ʂuo"],
+    "si": ["sɪ"],
+    "song": ["soŋ"],
+    "sou": ["sou"],
+    "su": ["su"],
+    "suan": ["suan"],
+    "sui": ["suei"],
+    "sun": ["sun"],
+    "suo": ["suo"],
+    "ta": ["ta"],
+    "tai": ["tai"],
+    "tan": ["tan"],
+    "tang": ["tɑŋ"],
+    "tao": ["taʌ"],
+    "te": ["tø"],
+    "tei": ["tei"],
+    "teng": ["tɵŋ"],
+    "ti": ["ti"],
+    "tian": ["tiɛn"],
+    "tiao": ["tiaʌ"],
+    "tie": ["tie"],
+    "ting": ["tɨŋ"],
+    "tong": ["toŋ"],
+    "tou": ["tou"],
+    "tu": ["tu"],
+    "tuan": ["tuan"],
+    "tui": ["tuei"],
+    "tun": ["tun"],
+    "tuo": ["tuo"],
+    "wa": ["wa"],
+    "wai": ["wai"],
+    "wan": ["wan"],
+    "wang": ["wɑŋ"],
+    "wei": ["wei"],
+    "wen": ["wœn"],
+    "weng": ["wɵŋ"],
+    "wo": ["wo"],
+    "wu": ["wu"],
+    "xi": ["ɕi"],
+    "xia": ["ɕia"],
+    "xian": ["ɕiɛn"],
+    "xiang": ["ɕiɑŋ"],
+    "xiao": ["ɕiaʌ"],
+    "xie": ["ɕie"],
+    "xin": ["ɕin"],
+    "xing": ["ɕɨŋ"],
+    "xiong": ["ɕioŋ"],
+    "xiu": ["ɕio"],
+    "xu": ["ɕy"],
+    "xuan": ["ɕyɛn"],
+    "xue": ["ɕye"],
+    "xun": ["ɕyn"],
+    "ya": ["ia"],
+    "yan": ["iɛn"],
+    "yang": ["iɑŋ"],
+    "yao": ["iaʌ"],
+    "ye": ["ie"],
+    "yi": ["i"],
+    "yin": ["in"],
+    "ying": ["ɨŋ"],
+    "yo": ["io"],
+    "yong": ["ioŋ"],
+    "you": ["io"],
+    "yu": ["y"],
+    "yuan": ["yɛn"], 
+    "yue": ["ye"],
+    "yun": ["yn"],
+    "za": ["dza"],
+    "zai": ["dzai"],
+    "zan": ["dzan"],
+    "zang": ["dzɑŋ"],
+    "zao": ["dzaʌ"],
+    "ze": ["dzø"],
+    "zei": ["dzei"],
+    "zen": ["dzœn"],
+    "zeng": ["dzɵŋ"],
+    "zha": ["dʒa"],
+    "zhai": ["dʒai"],
+    "zhan": ["dʒan"],
+    "zhang": ["dʒɑŋ"],
+    "zhao": ["dʒaʌ"],
+    "zhe": ["dʒø"],
+    # "zhei": ["dʒei"], it doesn't exist
+    "zhen": ["dʒœn"],
+    "zheng": ["dʒɵŋ"],
+    "zhi": ["dʒʏ"],
+    "zhong": ["dʒoŋ"],
+    "zhou": ["dʒou"],
+    "zhu": ["dʒu"],
+    "zhua": ["dʒua"],
+    "zhuai": ["dʒuai"],
+    "zhuan": ["dʒuan"],
+    "zhuang": ["dʒuɑŋ"],
+    "zhui": ["dʒuei"],
+    "zhun": ["dʒun"],
+    "zhuo": ["dʒuo"],
+    "zi": ["dzɪ"],
+    "zong": ["dzoŋ"],
+    "zou": ["dzou"],
+    "zu": ["dzu"],
+    "zuan": ["dzuan"],
+    "zui": ["dzuei"],
+    "zun": ["dzun"],
+    "zuo": ["dzuo"],
+}

TTS/tts/utils/synthesis.py

@@ -219,7 +219,7 @@ def synthesis(model,
             ap (TTS.tts.utils.audio.AudioProcessor): audio processor to process
                 model outputs.
             speaker_id (int): id of speaker
-            style_wav (str): Uses for style embedding of GST.
+            style_wav (str | Dict[str, float]): Uses for style embedding of GST.
             truncated (bool): keep model states after inference. It can be used
                 for continuous inference at long texts.
             enable_eos_bos_chars (bool): enable special chars for end of sentence and start of sentence.

TTS/tts/utils/text/__init__.py

@@ -8,6 +8,7 @@ from phonemizer.phonemize import phonemize
 from TTS.tts.utils.text import cleaners
 from TTS.tts.utils.text.symbols import (_bos, _eos, _punctuations,
                                         make_symbols, phonemes, symbols)
+from TTS.tts.utils.chinese_mandarin.phonemizer import chinese_text_to_phonemes
 
 
 # pylint: disable=unnecessary-comprehension
@@ -28,9 +29,23 @@ PHONEME_PUNCTUATION_PATTERN = r'['+_punctuations.replace(' ', '')+']+'
 
 
 def text2phone(text, language):
+    '''Convert graphemes to phonemes. For most of the languages, it calls
+    the phonemizer python library that calls espeak/espeak-ng. For chinese
+    mandarin, it calls pypinyin + custom function for phonemizing
+        Parameters:
+                text (str): text to phonemize
+                language (str): language of the text
+        Returns:
+                ph (str): phonemes as a string seperated by "|"
+                        ph = "ɪ|g|ˈ|z|æ|m|p|ə|l"
     '''
-    Convert graphemes to phonemes.
-    '''
+
+    # TO REVIEW : How to have a good implementation for this?
+    if language == "chinese-mandarin":
+        ph = chinese_text_to_phonemes(text)
+        return ph
+
+
     seperator = phonemizer.separator.Separator(' |', '', '|')
     #try:
     punctuations = re.findall(PHONEME_PUNCTUATION_PATTERN, text)

TTS/tts/utils/text/cleaners.py

@@ -15,6 +15,8 @@ from unidecode import unidecode
 from .number_norm import normalize_numbers
 from .abbreviations import abbreviations_en, abbreviations_fr
 from .time import expand_time_english
+from TTS.tts.utils.chinese_mandarin.numbers import replace_numbers_to_characters_in_text
+
 
 # Regular expression matching whitespace:
 _whitespace_re = re.compile(r'\s+')
@@ -122,6 +124,13 @@ def portuguese_cleaners(text):
     text = collapse_whitespace(text)
     return text
 
+def chinese_mandarin_cleaners(text: str) -> str:
+    '''Basic pipeline for chinese'''
+    text = replace_numbers_to_characters_in_text(text)
+    return text
+
+
+
 def phoneme_cleaners(text):
     '''Pipeline for phonemes mode, including number and abbreviation expansion.'''
     text = expand_numbers(text)

requirements.txt

@@ -6,6 +6,8 @@ numba==0.48
 librosa==0.7.2
 phonemizer>=2.2.0
 unidecode==0.4.20
+pypinyin
+jieba
 tensorboardX
 matplotlib
 Pillow