From 816798e8b594e852a204f34ed60ea9321dc40826 Mon Sep 17 00:00:00 2001
From: nadare <1na2da0re3@gmail.com>
Date: Sun, 9 Jul 2023 03:34:10 +0900
Subject: [PATCH] delete padding

---
 .../RVC/pitchExtractor/CrepeOnnxPitchExtractor.py             | 2 --
 .../voice_changer/RVC/pitchExtractor/CrepePitchExtractor.py   | 4 ----
 server/voice_changer/RVC/pitchExtractor/DioPitchExtractor.py  | 3 +--
 .../voice_changer/RVC/pitchExtractor/HarvestPitchExtractor.py | 2 --
 4 files changed, 1 insertion(+), 10 deletions(-)

diff --git a/server/voice_changer/RVC/pitchExtractor/CrepeOnnxPitchExtractor.py b/server/voice_changer/RVC/pitchExtractor/CrepeOnnxPitchExtractor.py
index fd126105..64b1ed42 100644
--- a/server/voice_changer/RVC/pitchExtractor/CrepeOnnxPitchExtractor.py
+++ b/server/voice_changer/RVC/pitchExtractor/CrepeOnnxPitchExtractor.py
@@ -54,8 +54,6 @@ class CrepeOnnxPitchExtractor(PitchExtractor):
         f0[pd < 0.1] = 0
         f0 = f0.squeeze()
 
-        f0 = np.pad(f0, (start_frame, n_frames - f0.shape[0] - start_frame), 'constant', constant_values=(0, 0))
-
         f0 *= pow(2, f0_up_key / 12)
         pitchf[-f0.shape[0]:] = f0[:pitchf.shape[0]]
         f0bak = pitchf.copy()
diff --git a/server/voice_changer/RVC/pitchExtractor/CrepePitchExtractor.py b/server/voice_changer/RVC/pitchExtractor/CrepePitchExtractor.py
index 2883e215..55528a0d 100644
--- a/server/voice_changer/RVC/pitchExtractor/CrepePitchExtractor.py
+++ b/server/voice_changer/RVC/pitchExtractor/CrepePitchExtractor.py
@@ -47,10 +47,6 @@ class CrepePitchExtractor(PitchExtractor):
         f0[pd < 0.1] = 0
         f0 = f0.squeeze()
 
-        f0 = torch.nn.functional.pad(
-            f0, (start_frame, n_frames - f0.shape[0] - start_frame)
-        )
-
         f0 *= pow(2, f0_up_key / 12)
         pitchf[-f0.shape[0]:] = f0.detach().cpu().numpy()[:pitchf.shape[0]]
         f0bak = pitchf.copy()
diff --git a/server/voice_changer/RVC/pitchExtractor/DioPitchExtractor.py b/server/voice_changer/RVC/pitchExtractor/DioPitchExtractor.py
index a0b0defb..a7d4c95b 100644
--- a/server/voice_changer/RVC/pitchExtractor/DioPitchExtractor.py
+++ b/server/voice_changer/RVC/pitchExtractor/DioPitchExtractor.py
@@ -17,7 +17,7 @@ class DioPitchExtractor(PitchExtractor):
         start_frame = int(silence_front * sr / window)
         real_silence_front = start_frame * window / sr
 
-        silence_front_offset = int(np.round(real_silence_front * sr))
+        silence_front_offset = max(min(int(np.round(real_silence_front * sr)), len(audio) - 3000), 0)
         audio = audio[silence_front_offset:]
 
         f0_min = 50
@@ -34,7 +34,6 @@ class DioPitchExtractor(PitchExtractor):
             frame_period=10,
         )
         f0 = pyworld.stonemask(audio.astype(np.double), _f0, t, sr)
-        # f0 = np.pad(f0.astype("float"), (start_frame, n_frames - len(f0) - start_frame))
 
         f0 *= pow(2, f0_up_key / 12)
         pitchf[-f0.shape[0]:] = f0[:pitchf.shape[0]]
diff --git a/server/voice_changer/RVC/pitchExtractor/HarvestPitchExtractor.py b/server/voice_changer/RVC/pitchExtractor/HarvestPitchExtractor.py
index 469799f0..776c0c2c 100644
--- a/server/voice_changer/RVC/pitchExtractor/HarvestPitchExtractor.py
+++ b/server/voice_changer/RVC/pitchExtractor/HarvestPitchExtractor.py
@@ -35,8 +35,6 @@ class HarvestPitchExtractor(PitchExtractor):
         f0 = pyworld.stonemask(audio.astype(np.double), f0, t, sr)
         f0 = signal.medfilt(f0, 3)
 
-        # f0 = np.pad(f0.astype("float"), (start_frame, n_frames - len(f0) - start_frame))
-
         f0 *= pow(2, f0_up_key / 12)
         pitchf[-f0.shape[0]:] = f0[:pitchf.shape[0]]
         f0bak = pitchf.copy()