From 37eaefc0852a9cf39fb325b658d67b9ee5758628 Mon Sep 17 00:00:00 2001
From: George <25833833+george-roussos@users.noreply.github.com>
Date: Fri, 29 Oct 2021 17:28:55 +0100
Subject: [PATCH] Optional silence trimming during inference and
 find_endpoint() fix (#898)

* Set find_endpoint db threshold in config.json

* Optional silence trimming during inference

* Make trim_db value negative
---
 TTS/utils/audio.py       | 4 ++--
 TTS/utils/synthesizer.py | 3 ++-
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/TTS/utils/audio.py b/TTS/utils/audio.py
index dd9c5701..e64b95e0 100644
--- a/TTS/utils/audio.py
+++ b/TTS/utils/audio.py
@@ -674,7 +674,7 @@ class AudioProcessor(object):
         return f0
 
     ### Audio Processing ###
-    def find_endpoint(self, wav: np.ndarray, threshold_db=-40, min_silence_sec=0.8) -> int:
+    def find_endpoint(self, wav: np.ndarray, min_silence_sec=0.8) -> int:
         """Find the last point without silence at the end of a audio signal.
 
         Args:
@@ -687,7 +687,7 @@ class AudioProcessor(object):
         """
         window_length = int(self.sample_rate * min_silence_sec)
         hop_length = int(window_length / 4)
-        threshold = self._db_to_amp(threshold_db)
+        threshold = self._db_to_amp(-self.trim_db)
         for x in range(hop_length, len(wav) - window_length, hop_length):
             if np.max(wav[x : x + window_length]) < threshold:
                 return x + hop_length
diff --git a/TTS/utils/synthesizer.py b/TTS/utils/synthesizer.py
index af07419f..78681967 100644
--- a/TTS/utils/synthesizer.py
+++ b/TTS/utils/synthesizer.py
@@ -265,7 +265,8 @@ class Synthesizer(object):
             waveform = waveform.squeeze()
 
             # trim silence
-            waveform = trim_silence(waveform, self.ap)
+            if self.tts_config.audio["do_trim_silence"] is True:
+                waveform = trim_silence(waveform, self.ap)
 
             wavs += list(waveform)
             wavs += [0] * 10000