From 1a40f9c21c3afd1e6b5a32471c497a9edc7c16b1 Mon Sep 17 00:00:00 2001
From: Binbin Zhang <binbzha@qq.com>
Date: Sun, 5 Nov 2023 09:02:01 +0800
Subject: [PATCH] normalize wav

---
 wespeaker/cli/speaker.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/wespeaker/cli/speaker.py b/wespeaker/cli/speaker.py
index 3c80a3af..fcb27c23 100644
--- a/wespeaker/cli/speaker.py
+++ b/wespeaker/cli/speaker.py
@@ -32,6 +32,7 @@ def __init__(self, model_path: str, resample_rate: int = 16000):
 
     def extract_embedding(self, audio_path: str):
         pcm, sample_rate = librosa.load(audio_path, sr=self.resample_rate)
+        pcm = pcm * (1 << 15)
         # NOTE: produce the same results as with torchaudio.compliance.kaldi
         feats = logfbank(
             pcm,