Handle espeak 1.48.15 (#2203)

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

@@ -1,6 +1,7 @@
 import logging
 import re
 import subprocess
+from distutils.version import LooseVersion
 from typing import Dict, List
 
 from TTS.tts.utils.text.phonemizers.base import BasePhonemizer
@@ -13,13 +14,26 @@ def is_tool(name):
     return which(name) is not None
 
 
+def get_espeak_version():
+    output = subprocess.getoutput("espeak --version")
+    return output.split()[2]
+
+
+def get_espeakng_version():
+    output = subprocess.getoutput("espeak-ng --version")
+    return output.split()[3]
+
+
 # priority: espeakng > espeak
 if is_tool("espeak-ng"):
     _DEF_ESPEAK_LIB = "espeak-ng"
+    _DEF_ESPEAK_VER = get_espeakng_version()
 elif is_tool("espeak"):
     _DEF_ESPEAK_LIB = "espeak"
+    _DEF_ESPEAK_VER = get_espeak_version()
 else:
     _DEF_ESPEAK_LIB = None
+    _DEF_ESPEAK_VER = None
 
 
 def _espeak_exe(espeak_lib: str, args: List, sync=False) -> List[str]:
@@ -85,6 +99,7 @@ class ESpeak(BasePhonemizer):
     """
 
     _ESPEAK_LIB = _DEF_ESPEAK_LIB
+    _ESPEAK_VER = _DEF_ESPEAK_VER
 
     def __init__(self, language: str, backend=None, punctuations=Punctuation.default_puncs(), keep_puncs=True):
         if self._ESPEAK_LIB is None:
@@ -105,17 +120,24 @@ class ESpeak(BasePhonemizer):
     def backend(self):
         return self._ESPEAK_LIB
 
+    @property
+    def backend_version(self):
+        return self._ESPEAK_VER
+
     @backend.setter
     def backend(self, backend):
         if backend not in ["espeak", "espeak-ng"]:
             raise Exception("Unknown backend: %s" % backend)
         self._ESPEAK_LIB = backend
+        self._ESPEAK_VER = get_espeakng_version() if backend == "espeak-ng" else get_espeak_version()
 
     def auto_set_espeak_lib(self) -> None:
         if is_tool("espeak-ng"):
             self._ESPEAK_LIB = "espeak-ng"
+            self._ESPEAK_VER = get_espeakng_version()
         elif is_tool("espeak"):
             self._ESPEAK_LIB = "espeak"
+            self._ESPEAK_VER = get_espeak_version()
         else:
             raise Exception("Cannot set backend automatically. espeak-ng or espeak not found")
 
@@ -146,7 +168,10 @@ class ESpeak(BasePhonemizer):
         else:
             # split with '_'
             if self.backend == "espeak":
-                args.append("--ipa=3")
+                if LooseVersion(self.backend_version) >= LooseVersion("1.48.15"):
+                    args.append("--ipa=1")
+                else:
+                    args.append("--ipa=3")
             else:
                 args.append("--ipa=1")
         if tie:

tests/text_tests/test_phonemizer.py

@@ -1,4 +1,5 @@
 import unittest
+from distutils.version import LooseVersion
 
 from TTS.tts.utils.text.phonemizers import ESpeak, Gruut, JA_JP_Phonemizer, ZH_CN_Phonemizer
 
@@ -18,6 +19,14 @@ EXPECTED_ESPEAK_PHONEMES = [
 ]
 
 
+EXPECTED_ESPEAK_v1_48_15_PHONEMES = [
+    "ɹ|ˈiː|s|ə|n|t ɹ|ɪ|s|ˈɜː|tʃ æ|t h|ˈɑːɹ|v|ɚ|d h|ɐ|z ʃ|ˈoʊ|n m|ˈɛ|d|ᵻ|t|ˌeɪ|ɾ|ɪ|ŋ",
+    "f|ɔː|ɹ æ|z l|ˈɪ|ɾ|əl æ|z ˈeɪ|t w|ˈiː|k|s k|æ|n ˈæ|k|tʃ|uː|əl|i| ˈɪ|n|k|ɹ|iː|s, ð|ə ɡ|ɹ|ˈeɪ m|ˈæ|ɾ|ɚ",
+    "ɪ|n|ð|ə p|ˈɑːɹ|t|s ʌ|v|ð|ə b|ɹ|ˈeɪ|n ɹ|ɪ|s|p|ˈɑː|n|s|ə|b|əl",
+    "f|ɔː|ɹ ɪ|m|ˈoʊ|ʃ|ə|n|əl ɹ|ˌɛ|ɡ|j|uː|l|ˈeɪ|ʃ|ə|n|| æ|n|d l|ˈɜː|n|ɪ|ŋ!",
+]
+
+
 EXPECTED_ESPEAKNG_PHONEMES = [
     "ɹ|ˈiː|s|ə|n|t ɹ|ᵻ|s|ˈɜː|tʃ æ|t h|ˈɑːɹ|v|ɚ|d h|ɐ|z ʃ|ˈoʊ|n m|ˈɛ|d|ᵻ|t|ˌeɪ|ɾ|ɪ|ŋ",
     "f|ɔː|ɹ æ|z l|ˈɪ|ɾ|əl æ|z ˈeɪ|t w|ˈiː|k|s k|æ|n ˈæ|k|tʃ|uː|əl|i| ˈɪ|ŋ|k|ɹ|iː|s, ð|ə ɡ|ɹ|ˈeɪ m|ˈæ|ɾ|ɚ",
@@ -30,13 +39,20 @@ class TestEspeakPhonemizer(unittest.TestCase):
     def setUp(self):
         self.phonemizer = ESpeak(language="en-us", backend="espeak")
 
-        for text, ph in zip(EXAMPLE_TEXTs, EXPECTED_ESPEAK_PHONEMES):
+        if LooseVersion(self.phonemizer.backend_version) >= LooseVersion("1.48.15"):
+            target_phonemes = EXPECTED_ESPEAK_v1_48_15_PHONEMES
+        else:
+            target_phonemes = EXPECTED_ESPEAK_PHONEMES
+
+        for text, ph in zip(EXAMPLE_TEXTs, target_phonemes):
             phonemes = self.phonemizer.phonemize(text)
             self.assertEqual(phonemes, ph)
 
         # multiple punctuations
         text = "Be a voice, not an! echo?"
         gt = "biː ɐ vˈɔɪs, nˈɑːt ɐn! ˈɛkoʊ?"
+        if LooseVersion(self.phonemizer.backend_version) >= LooseVersion("1.48.15"):
+            gt = "biː ɐ vˈɔɪs, nˈɑːt æn! ˈɛkoʊ?"
         output = self.phonemizer.phonemize(text, separator="|")
         output = output.replace("|", "")
         self.assertEqual(output, gt)
@@ -44,12 +60,16 @@ class TestEspeakPhonemizer(unittest.TestCase):
         # not ending with punctuation
         text = "Be a voice, not an! echo"
         gt = "biː ɐ vˈɔɪs, nˈɑːt ɐn! ˈɛkoʊ"
+        if LooseVersion(self.phonemizer.backend_version) >= LooseVersion("1.48.15"):
+            gt = "biː ɐ vˈɔɪs, nˈɑːt æn! ˈɛkoʊ"
         output = self.phonemizer.phonemize(text, separator="")
         self.assertEqual(output, gt)
 
         # extra space after the sentence
         text = "Be a voice, not an! echo.  "
         gt = "biː ɐ vˈɔɪs, nˈɑːt ɐn! ˈɛkoʊ."
+        if LooseVersion(self.phonemizer.backend_version) >= LooseVersion("1.48.15"):
+            gt = "biː ɐ vˈɔɪs, nˈɑːt æn! ˈɛkoʊ."
         output = self.phonemizer.phonemize(text, separator="")
         self.assertEqual(output, gt)