From 3c86cd5bca03393da14a7c94b1ef3d700962b880 Mon Sep 17 00:00:00 2001
From: babysor00 <babysor00@gmail.com>
Date: Sun, 22 Aug 2021 23:44:25 +0800
Subject: [PATCH] =?UTF-8?q?=E3=80=90bugfix=E3=80=91=20fix=20bug=20causing?=
 =?UTF-8?q?=20non-sense=20output=20for=20long=20texts=20=E4=BF=AE=E5=A4=8D?=
 =?UTF-8?q?=E5=A4=9A=E6=AE=B5=E6=96=87=E5=AD=97=E5=8F=91=E9=9F=B3=E9=94=99?=
 =?UTF-8?q?=E8=AF=AF?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 synthesizer/inference.py     | 7 ++-----
 synthesizer/utils/symbols.py | 2 +-
 toolbox/__init__.py          | 8 ++++++++
 3 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/synthesizer/inference.py b/synthesizer/inference.py
index 07cf881..694134f 100644
--- a/synthesizer/inference.py
+++ b/synthesizer/inference.py
@@ -90,13 +90,10 @@ class Synthesizer:
 
             simple_table([("Tacotron", str(tts_k) + "k"),
                         ("r", self._model.r)])
-
-        #convert chinese char to pinyin
-        list_of_pinyin = lazy_pinyin(texts, style=Style.TONE3)
-        texts = [" ".join([v for v in list_of_pinyin if v.strip()])]
+        texts = [" ".join(lazy_pinyin(v, style=Style.TONE3)) for v in texts]
 
         # Preprocess text inputs
-        inputs = [text_to_sequence(text.strip(), hparams.tts_cleaner_names) for text in texts]
+        inputs = [text_to_sequence(text, hparams.tts_cleaner_names) for text in texts]
         if not isinstance(embeddings, list):
             embeddings = [embeddings]
 
diff --git a/synthesizer/utils/symbols.py b/synthesizer/utils/symbols.py
index 2dbec9a..d9c3967 100644
--- a/synthesizer/utils/symbols.py
+++ b/synthesizer/utils/symbols.py
@@ -8,7 +8,7 @@ through Unidecode. For other data, you can modify _characters. See TRAINING_DATA
 
 _pad        = "_"
 _eos        = "~"
-_characters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz12340!\'(),-.:;? '
+_characters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz1234567890!\'(),-.:;? '
 # Prepend "@" to ARPAbet symbols to ensure uniqueness (some are the same as uppercase letters):
 #_arpabet = ["@' + s for s in cmudict.valid_symbols]
 
diff --git a/toolbox/__init__.py b/toolbox/__init__.py
index 74c6b3c..22eda81 100644
--- a/toolbox/__init__.py
+++ b/toolbox/__init__.py
@@ -10,6 +10,7 @@ import traceback
 import sys
 import torch
 import librosa
+import re
 from audioread.exceptions import NoBackendError
 
 # Use this directory structure for your datasets, or modify it to fit your needs
@@ -224,6 +225,13 @@ class Toolbox:
             self.init_synthesizer()
 
         texts = self.ui.text_prompt.toPlainText().split("\n")
+        punctuation = '！，。、,' # punctuate and split/clean text
+        processed_texts = []
+        for text in texts:
+          for processed_text in re.sub(r'[{}]+'.format(punctuation), '\n', text).split('\n'):
+            if processed_text:
+                processed_texts.append(processed_text.strip())
+        texts = processed_texts
         embed = self.ui.selected_utterance.embed
         embeds = [embed] * len(texts)
         specs = self.synthesizer.synthesize_spectrograms(texts, embeds)