From 07cbcf825c8837838229bb1b6952bd35191133eb Mon Sep 17 00:00:00 2001
From: Enno Hermann <enno.hermann@idiap.ch>
Date: Wed, 29 May 2024 09:52:18 +0200
Subject: [PATCH] fix(espeak_wrapper): read phonemize() input from file

Avoids utf8 encoding issues on Windows when passing the text directly.
Fixes https://github.com/coqui-ai/TTS/discussions/3761
---
 TTS/tts/utils/text/phonemizers/espeak_wrapper.py | 10 +++++++++-
 tests/text_tests/test_phonemizer.py              |  6 ++++++
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/TTS/tts/utils/text/phonemizers/espeak_wrapper.py b/TTS/tts/utils/text/phonemizers/espeak_wrapper.py
index 91fb93c7..a15df716 100644
--- a/TTS/tts/utils/text/phonemizers/espeak_wrapper.py
+++ b/TTS/tts/utils/text/phonemizers/espeak_wrapper.py
@@ -3,6 +3,8 @@
 import logging
 import re
 import subprocess
+import tempfile
+from pathlib import Path
 from typing import Optional
 
 from packaging.version import Version
@@ -184,7 +186,12 @@ class ESpeak(BasePhonemizer):
         if tie:
             args.append("--tie=%s" % tie)
 
-        args.append(text)
+        tmp = tempfile.NamedTemporaryFile(mode="w+t", delete=False, encoding="utf8")
+        tmp.write(text)
+        tmp.close()
+        args.append("-f")
+        args.append(tmp.name)
+
         # compute phonemes
         phonemes = ""
         for line in _espeak_exe(self.backend, args):
@@ -200,6 +207,7 @@ class ESpeak(BasePhonemizer):
             ph_decoded = re.sub(r"\(.+?\)", "", line)
 
             phonemes += ph_decoded.strip()
+        Path(tmp.name).unlink()
         return phonemes.replace("_", separator)
 
     def _phonemize(self, text: str, separator: str = "") -> str:
diff --git a/tests/text_tests/test_phonemizer.py b/tests/text_tests/test_phonemizer.py
index ca25b302..f9067530 100644
--- a/tests/text_tests/test_phonemizer.py
+++ b/tests/text_tests/test_phonemizer.py
@@ -116,6 +116,12 @@ class TestEspeakNgPhonemizer(unittest.TestCase):
         output = self.phonemizer.phonemize(text, separator="")
         self.assertEqual(output, gt)
 
+        # UTF8 characters
+        text = "źrebię"
+        gt = "ʑrˈɛbjɛ"
+        output = ESpeak("pl").phonemize(text, separator="")
+        self.assertEqual(output, gt)
+
     def test_name(self):
         self.assertEqual(self.phonemizer.name(), "espeak")