From 41edcf899dc7eb7e68e8a136f872ff6d7b6d93c2 Mon Sep 17 00:00:00 2001
From: coolBrown12 <yash91238@gmail.com>
Date: Sat, 13 May 2023 23:51:13 +0530
Subject: [PATCH] This code defines a function preprocess_split that takes an
 input directory of audio files, processes each file by splitting it into
 segments based on silence, and saves the resulting segments to an output
 directory.

---
 .../preprocessing/tempCodeRunnerFile.py       | 70 +++++++++++++++++++
 1 file changed, 70 insertions(+)
 create mode 100644 src/so_vits_svc_fork/preprocessing/tempCodeRunnerFile.py

diff --git a/src/so_vits_svc_fork/preprocessing/tempCodeRunnerFile.py b/src/so_vits_svc_fork/preprocessing/tempCodeRunnerFile.py
new file mode 100644
index 00000000..fb98938a
--- /dev/null
+++ b/src/so_vits_svc_fork/preprocessing/tempCodeRunnerFile.py
@@ -0,0 +1,70 @@
+from __future__ import annotations
+
+from logging import getLogger
+from pathlib import Path
+
+import librosa
+import soundfile as sf
+from joblib import Parallel, delayed
+from tqdm import tqdm
+from tqdm_joblib import tqdm_joblib
+
+LOG = getLogger(__name__)
+
+
+def _process_one(
+    input_path: Path,
+    output_dir: Path,
+    sr: int,
+    *,
+    top_db: int = 30,
+    frame_seconds: float = 0.5,
+    hop_seconds: float = 0.1,
+):
+    try:
+        audio, sr = librosa.load(input_path, sr=sr, mono=True)
+    except Exception as e:
+        LOG.warning(f"Failed to read {input_path}: {e}")
+        return
+    intervals = librosa.effects.split(
+        audio,
+        top_db=top_db,
+        frame_length=int(sr * frame_seconds),
+        hop_length=int(sr * hop_seconds),
+    )
+    output_dir.mkdir(parents=True, exist_ok=True)
+    for start, end in tqdm(intervals, desc=f"Writing {input_path}"):
+        audio_cut = audio[start:end]
+        sf.write(
+            (output_dir / f"{input_path.stem}_{start / sr:.3f}_{end / sr:.3f}.wav"),
+            audio_cut,
+            sr,
+        )
+
+
+def preprocess_split(
+    input_dir: Path | str,
+    output_dir: Path | str,
+    sr: int,
+    *,
+    top_db: int = 30,
+    frame_seconds: float = 0.5,
+    hop_seconds: float = 0.1,
+    n_jobs: int = -1,
+):
+    input_dir = Path(input_dir)
+    output_dir = Path(output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+    input_paths = list(input_dir.rglob("*.*"))
+    with tqdm_joblib(desc="Splitting", total=len(input_paths)):
+        Parallel(n_jobs=n_jobs)(
+            delayed(_process_one)(
+                input_path,
+                output_dir / input_path.relative_to(input_dir).parent,
+                sr,
+                top_db=top_db,
+                frame_seconds=frame_seconds,
+                hop_seconds=hop_seconds,
+            )
+            for input_path in input_paths
+        )