From 60034674f9696d79e1683b958b3f1fdb43412b1c Mon Sep 17 00:00:00 2001 From: Edresson Casanova Date: Thu, 28 Apr 2022 15:40:16 -0300 Subject: [PATCH] Remove audio padding before mel spec extraction --- TTS/vocoder/datasets/gan_dataset.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/TTS/vocoder/datasets/gan_dataset.py b/TTS/vocoder/datasets/gan_dataset.py index 08fdb0a3cf..a782067e1b 100644 --- a/TTS/vocoder/datasets/gan_dataset.py +++ b/TTS/vocoder/datasets/gan_dataset.py @@ -115,8 +115,8 @@ def load_item(self, idx): audio, mel = self.cache[idx] else: audio = self.ap.load_wav(wavpath) - audio, _ = self._pad_short_samples(audio) mel = self.ap.melspectrogram(audio) + audio, mel = self._pad_short_samples(audio, mel) else: # load precomputed features