fix: fix audio resampled to 22kHz (#130)

src/so_vits_svc_fork/gui.py

@@ -4,9 +4,9 @@
 from logging import getLogger
 from pathlib import Path
 
-import librosa
 import PySimpleGUI as sg
 import sounddevice as sd
+import soundfile as sf
 import torch
 from pebble import ProcessFuture, ProcessPool
 from tqdm.tk import tqdm_tk
@@ -24,7 +24,7 @@
 def play_audio(path: Path | str):
     if isinstance(path, Path):
         path = path.as_posix()
-    data, sr = librosa.load(path)
+    data, sr = sf.read(path)
     sd.play(data, sr)
 
 

src/so_vits_svc_fork/preprocess_resample.py

@@ -9,6 +9,7 @@
 import librosa
 import numpy as np
 import soundfile
+import soundfile as sf
 from joblib import Parallel, delayed
 from tqdm_joblib import tqdm_joblib
 
@@ -58,7 +59,7 @@ def preprocess_one(input_path: Path, output_path: Path) -> None:
         """Preprocess one audio file."""
 
         try:
-            audio, sr = librosa.load(input_path)
+            audio, sr = sf.read(input_path)
 
         # Audioread is the last backend it will attempt, so this is the exception thrown on failure
         except audioread.exceptions.NoBackendError as e:

src/so_vits_svc_fork/preprocess_split.py

@@ -2,6 +2,7 @@
 from pathlib import Path
 
 import librosa
+import soundfile
 import soundfile as sf
 from joblib import Parallel, delayed
 from tqdm import tqdm
@@ -19,7 +20,7 @@ def _process_one(
     hop_seconds: float = 0.1,
 ):
     try:
-        audio, sr = librosa.load(input_path)
+        audio, sr = sf.read(input_path)
     except Exception as e:
         LOG.warning(f"Failed to read {input_path}: {e}")
         return