mirror of
https://github.com/RVC-Boss/GPT-SoVITS.git
synced 2025-10-09 00:10:00 +08:00
479 lines
15 KiB
Python
479 lines
15 KiB
Python
# modified from https://github.com/fishaudio/audio-preprocess/blob/main/fish_audio_preprocess/cli/slice_audio.py
|
|
|
|
from pathlib import Path
|
|
from typing import Union, Iterable
|
|
from concurrent.futures import ProcessPoolExecutor, as_completed
|
|
from tqdm import tqdm
|
|
import librosa
|
|
import numpy as np
|
|
import soundfile as sf
|
|
import math
|
|
from my_utils import load_audio
|
|
import argparse
|
|
import os
|
|
|
|
|
|
AUDIO_EXTENSIONS = {
|
|
".mp3",
|
|
".wav",
|
|
".flac",
|
|
".ogg",
|
|
".m4a",
|
|
".wma",
|
|
".aac",
|
|
".aiff",
|
|
".aif",
|
|
".aifc",
|
|
}
|
|
|
|
|
|
def list_files(
|
|
path: Union[Path, str],
|
|
extensions: set[str] = None,
|
|
sort: bool = True,
|
|
) -> list[Path]:
|
|
"""List files in a directory.
|
|
|
|
Args:
|
|
path (Path): Path to the directory.
|
|
extensions (set, optional): Extensions to filter. Defaults to None.
|
|
sort (bool, optional): Whether to sort the files. Defaults to True.
|
|
|
|
Returns:
|
|
list: List of files.
|
|
"""
|
|
|
|
if isinstance(path, str):
|
|
path = Path(path)
|
|
|
|
if not path.exists():
|
|
raise FileNotFoundError(f"Directory {path} does not exist.")
|
|
|
|
files = [f for f in path.glob("*") if f.is_file()]
|
|
|
|
if extensions is not None:
|
|
files = [f for f in files if f.suffix in extensions]
|
|
|
|
if sort:
|
|
files = sorted(files)
|
|
|
|
return files
|
|
|
|
|
|
def make_dirs(path: Union[Path, str]):
|
|
"""Make directories.
|
|
|
|
Args:
|
|
path (Union[Path, str]): Path to the directory.
|
|
"""
|
|
if isinstance(path, str):
|
|
path = Path(path)
|
|
|
|
if path.exists():
|
|
print(f"Output directory already exists: {path}")
|
|
|
|
path.mkdir(parents=True, exist_ok=True)
|
|
|
|
|
|
def slice_audio_v2_(
|
|
input_path: str,
|
|
output_dir: str,
|
|
num_workers: int,
|
|
min_duration: float,
|
|
max_duration: float,
|
|
min_silence_duration: float,
|
|
top_db: int,
|
|
hop_length: int,
|
|
max_silence_kept: float,
|
|
merge_short:bool
|
|
):
|
|
"""(OpenVPI version) Slice audio files into smaller chunks by silence."""
|
|
|
|
input_path_, output_dir_ = Path(input_path), Path(output_dir)
|
|
if not input_path_.exists():
|
|
raise RuntimeError("You input a wrong audio path that does not exists, please fix it!")
|
|
make_dirs(output_dir_)
|
|
if input_path_.is_dir():
|
|
files = list_files(input_path_, extensions=AUDIO_EXTENSIONS)
|
|
elif input_path_.is_file() and input_path_.suffix in AUDIO_EXTENSIONS:
|
|
files = [input_path_]
|
|
input_path_ = input_path_.parent
|
|
else:
|
|
raise RuntimeError("The input path is not file or dir, please fixes it")
|
|
print(f"Found {len(files)} files, processing...")
|
|
|
|
|
|
with ProcessPoolExecutor(max_workers=num_workers) as executor:
|
|
tasks = []
|
|
|
|
for file in tqdm(files, desc="Preparing tasks"):
|
|
# Get relative path to input_dir
|
|
relative_path = file.relative_to(input_path_)
|
|
save_path = output_dir_ / relative_path.parent / relative_path.stem
|
|
|
|
tasks.append(
|
|
executor.submit(
|
|
slice_audio_file_v2,
|
|
input_file=str(file),
|
|
output_dir=save_path,
|
|
min_duration=min_duration,
|
|
max_duration=max_duration,
|
|
min_silence_duration=min_silence_duration,
|
|
top_db=top_db,
|
|
hop_length=hop_length,
|
|
max_silence_kept=max_silence_kept,
|
|
merge_short=merge_short
|
|
)
|
|
)
|
|
|
|
for i in tqdm(as_completed(tasks), total=len(tasks), desc="Processing"):
|
|
assert i.exception() is None, i.exception()
|
|
|
|
print("Done!")
|
|
print(f"Total: {len(files)}")
|
|
print(f"Output directory: {output_dir}")
|
|
|
|
|
|
def slice_audio_file_v2(
|
|
input_file: Union[str, Path],
|
|
output_dir: Union[str, Path],
|
|
min_duration: float = 4.0,
|
|
max_duration: float = 12.0,
|
|
min_silence_duration: float = 0.3,
|
|
top_db: int = -40,
|
|
hop_length: int = 10,
|
|
max_silence_kept: float = 0.4,
|
|
merge_short: bool = False
|
|
) -> None:
|
|
"""
|
|
Slice audio by silence and save to output folder
|
|
|
|
Args:
|
|
input_file: input audio file
|
|
output_dir: output folder
|
|
min_duration: minimum duration of each slice
|
|
max_duration: maximum duration of each slice
|
|
min_silence_duration: minimum duration of silence
|
|
top_db: threshold to detect silence
|
|
hop_length: hop length to detect silence
|
|
max_silence_kept: maximum duration of silence to be kept
|
|
"""
|
|
|
|
output_dir = Path(output_dir)
|
|
|
|
audio = load_audio(str(input_file),32000)
|
|
rate = 32000
|
|
for idx, sliced in enumerate(
|
|
slice_audio_v2(
|
|
audio,
|
|
rate,
|
|
min_duration=min_duration,
|
|
max_duration=max_duration,
|
|
min_silence_duration=min_silence_duration,
|
|
top_db=top_db,
|
|
hop_length=hop_length,
|
|
max_silence_kept=max_silence_kept,
|
|
merge_short=merge_short
|
|
)
|
|
):
|
|
if len(sliced) <= 3*rate: continue
|
|
max_audio=np.abs(sliced).max()#防止爆音,懒得搞混合了,后面有响度匹配
|
|
if max_audio>1:
|
|
sliced/=max_audio
|
|
sf.write(str(output_dir) + f"_{idx:04d}.wav", sliced, rate)
|
|
|
|
|
|
def slice_audio_v2(
|
|
audio: np.ndarray,
|
|
rate: int,
|
|
min_duration: float = 4.0,
|
|
max_duration: float = 12.0,
|
|
min_silence_duration: float = 0.3,
|
|
top_db: int = -40,
|
|
hop_length: int = 10,
|
|
max_silence_kept: float = 0.5,
|
|
merge_short: bool = False
|
|
) -> Iterable[np.ndarray]:
|
|
"""Slice audio by silence
|
|
|
|
Args:
|
|
audio: audio data, in shape (samples, channels)
|
|
rate: sample rate
|
|
min_duration: minimum duration of each slice
|
|
max_duration: maximum duration of each slice
|
|
min_silence_duration: minimum duration of silence
|
|
top_db: threshold to detect silence
|
|
hop_length: hop length to detect silence
|
|
max_silence_kept: maximum duration of silence to be kept
|
|
merge_short: merge short slices automatically
|
|
|
|
Returns:
|
|
Iterable of sliced audio
|
|
"""
|
|
|
|
if len(audio) / rate < min_duration:
|
|
sliced_by_max_duration_chunk = slice_by_max_duration(audio, max_duration, rate)
|
|
yield from merge_short_chunks(
|
|
sliced_by_max_duration_chunk, max_duration, rate
|
|
) if merge_short else sliced_by_max_duration_chunk
|
|
return
|
|
|
|
slicer = Slicer(
|
|
sr=rate,
|
|
threshold=top_db,
|
|
min_length=min_duration * 1000,
|
|
min_interval=min_silence_duration * 1000,
|
|
hop_size=hop_length,
|
|
max_sil_kept=max_silence_kept * 1000,
|
|
)
|
|
|
|
sliced_audio = slicer.slice(audio)
|
|
if merge_short:
|
|
sliced_audio = merge_short_chunks(sliced_audio, max_duration, rate)
|
|
|
|
for chunk in sliced_audio:
|
|
sliced_by_max_duration_chunk = slice_by_max_duration(chunk, max_duration, rate)
|
|
yield from sliced_by_max_duration_chunk
|
|
|
|
|
|
def slice_by_max_duration(
|
|
gen: np.ndarray, slice_max_duration: float, rate: int
|
|
) -> Iterable[np.ndarray]:
|
|
"""Slice audio by max duration
|
|
|
|
Args:
|
|
gen: audio data, in shape (samples, channels)
|
|
slice_max_duration: maximum duration of each slice
|
|
rate: sample rate
|
|
|
|
Returns:
|
|
generator of sliced audio data
|
|
"""
|
|
|
|
if len(gen) > slice_max_duration * rate:
|
|
# Evenly split _gen into multiple slices
|
|
n_chunks = math.ceil(len(gen) / (slice_max_duration * rate))
|
|
chunk_size = math.ceil(len(gen) / n_chunks)
|
|
|
|
for i in range(0, len(gen), chunk_size):
|
|
yield gen[i : i + chunk_size]
|
|
else:
|
|
yield gen
|
|
|
|
|
|
class Slicer:
|
|
def __init__(
|
|
self,
|
|
sr: int,
|
|
threshold: float = -40.0,
|
|
min_length: int = 4000,
|
|
min_interval: int = 300,
|
|
hop_size: int = 10,
|
|
max_sil_kept: int = 5000,
|
|
):
|
|
if not min_length >= min_interval >= hop_size:
|
|
raise ValueError(
|
|
"The following condition must be satisfied: min_length >= min_interval >= hop_size"
|
|
)
|
|
|
|
if not max_sil_kept >= hop_size:
|
|
raise ValueError(
|
|
"The following condition must be satisfied: max_sil_kept >= hop_size"
|
|
)
|
|
|
|
min_interval = sr * min_interval / 1000
|
|
self.threshold = 10 ** (threshold / 20.0)
|
|
self.hop_size = round(sr * hop_size / 1000)
|
|
self.win_size = min(round(min_interval), 4 * self.hop_size)
|
|
self.min_length = round(sr * min_length / 1000 / self.hop_size)
|
|
self.min_interval = round(min_interval / self.hop_size)
|
|
self.max_sil_kept = round(sr * max_sil_kept / 1000 / self.hop_size)
|
|
|
|
def _apply_slice(self, waveform, begin, end):
|
|
if len(waveform.shape) > 1:
|
|
return waveform[
|
|
:, begin * self.hop_size : min(waveform.shape[1], end * self.hop_size)
|
|
]
|
|
else:
|
|
return waveform[
|
|
begin * self.hop_size : min(waveform.shape[0], end * self.hop_size)
|
|
]
|
|
|
|
def slice(self, waveform):
|
|
if len(waveform.shape) > 1:
|
|
samples = waveform.mean(axis=0)
|
|
else:
|
|
samples = waveform
|
|
|
|
if samples.shape[0] <= self.min_length:
|
|
return [waveform]
|
|
|
|
rms_list = librosa.feature.rms(
|
|
y=samples, frame_length=self.win_size, hop_length=self.hop_size
|
|
).squeeze(0)
|
|
sil_tags = []
|
|
silence_start = None
|
|
clip_start = 0
|
|
|
|
for i, rms in enumerate(rms_list):
|
|
# Keep looping while frame is silent.
|
|
if rms < self.threshold:
|
|
# Record start of silent frames.
|
|
if silence_start is None:
|
|
silence_start = i
|
|
continue
|
|
|
|
# Keep looping while frame is not silent and silence start has not been recorded.
|
|
if silence_start is None:
|
|
continue
|
|
|
|
# Clear recorded silence start if interval is not enough or clip is too short
|
|
is_leading_silence = silence_start == 0 and i > self.max_sil_kept
|
|
need_slice_middle = (
|
|
i - silence_start >= self.min_interval
|
|
and i - clip_start >= self.min_length
|
|
)
|
|
|
|
if not is_leading_silence and not need_slice_middle:
|
|
silence_start = None
|
|
continue
|
|
|
|
# Need slicing. Record the range of silent frames to be removed.
|
|
if i - silence_start <= self.max_sil_kept:
|
|
pos = rms_list[silence_start : i + 1].argmin() + silence_start
|
|
|
|
if silence_start == 0:
|
|
sil_tags.append((0, pos))
|
|
else:
|
|
sil_tags.append((pos, pos))
|
|
|
|
clip_start = pos
|
|
elif i - silence_start <= self.max_sil_kept * 2:
|
|
pos = rms_list[
|
|
i - self.max_sil_kept : silence_start + self.max_sil_kept + 1
|
|
].argmin()
|
|
pos += i - self.max_sil_kept
|
|
pos_l = (
|
|
rms_list[
|
|
silence_start : silence_start + self.max_sil_kept + 1
|
|
].argmin()
|
|
+ silence_start
|
|
)
|
|
pos_r = (
|
|
rms_list[i - self.max_sil_kept : i + 1].argmin()
|
|
+ i
|
|
- self.max_sil_kept
|
|
)
|
|
|
|
if silence_start == 0:
|
|
sil_tags.append((0, pos_r))
|
|
clip_start = pos_r
|
|
else:
|
|
sil_tags.append((min(pos_l, pos), max(pos_r, pos)))
|
|
clip_start = max(pos_r, pos)
|
|
else:
|
|
pos_l = (
|
|
rms_list[
|
|
silence_start : silence_start + self.max_sil_kept + 1
|
|
].argmin()
|
|
+ silence_start
|
|
)
|
|
pos_r = (
|
|
rms_list[i - self.max_sil_kept : i + 1].argmin()
|
|
+ i
|
|
- self.max_sil_kept
|
|
)
|
|
|
|
if silence_start == 0:
|
|
sil_tags.append((0, pos_r))
|
|
else:
|
|
sil_tags.append((pos_l, pos_r))
|
|
|
|
clip_start = pos_r
|
|
silence_start = None
|
|
|
|
# Deal with trailing silence.
|
|
total_frames = rms_list.shape[0]
|
|
if (
|
|
silence_start is not None
|
|
and total_frames - silence_start >= self.min_interval
|
|
):
|
|
silence_end = min(total_frames, silence_start + self.max_sil_kept)
|
|
pos = rms_list[silence_start : silence_end + 1].argmin() + silence_start
|
|
sil_tags.append((pos, total_frames + 1))
|
|
|
|
# Apply and return slices.
|
|
if len(sil_tags) == 0:
|
|
return [waveform]
|
|
else:
|
|
chunks = []
|
|
|
|
if sil_tags[0][0] > 0:
|
|
chunks.append(self._apply_slice(waveform, 0, sil_tags[0][0]))
|
|
|
|
for i in range(len(sil_tags) - 1):
|
|
chunks.append(
|
|
self._apply_slice(waveform, sil_tags[i][1], sil_tags[i + 1][0])
|
|
)
|
|
|
|
if sil_tags[-1][1] < total_frames:
|
|
chunks.append(
|
|
self._apply_slice(waveform, sil_tags[-1][1], total_frames)
|
|
)
|
|
|
|
return chunks
|
|
|
|
|
|
def merge_short_chunks(chunks, max_duration, rate):
|
|
merged_chunks = []
|
|
buffer, length = [], 0
|
|
|
|
for chunk in chunks:
|
|
if length + len(chunk) > max_duration * rate and len(buffer) > 0:
|
|
merged_chunks.append(np.concatenate(buffer))
|
|
buffer, length = [], 0
|
|
else:
|
|
buffer.append(chunk)
|
|
length += len(chunk)
|
|
|
|
if len(buffer) > 0:
|
|
merged_chunks.append(np.concatenate(buffer))
|
|
|
|
return merged_chunks
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
parser = argparse.ArgumentParser()
|
|
parser.add_argument("-i","--input_dir",help="切割输入文件夹")
|
|
parser.add_argument("-o","--output_dir",help="切割输入文件夹")
|
|
parser.add_argument("--threshold",default=-40,help="音量小于这个值视作静音的备选切割点")
|
|
parser.add_argument("--min_duration",default=4,help="每段最短多长，如果第一段太短一直和后面段连起来直到超过这个值")
|
|
parser.add_argument("--max_duration",default=12,help="每段最长多长")
|
|
parser.add_argument("--min_interval",default=0.3,help="最短切割间隔")
|
|
parser.add_argument("--hop_size",default=10,help="怎么算音量曲线，越小精度越大计算量越高（不是精度越大效果越好）")
|
|
parser.add_argument("--max_sil_kept",default=0.4,help="切完后静音最多留多长")
|
|
parser.add_argument("--num_worker",default=os.cpu_count(),help="切使用的进程数")
|
|
parser.add_argument("--merge_short",default="False",help="响割使用的进程数")
|
|
|
|
|
|
args = parser.parse_args()
|
|
input_path = args.input_dir
|
|
output_dir = args.output_dir
|
|
threshold = float(args.threshold)
|
|
min_duration = float(args.min_duration)
|
|
max_duration = float(args.max_duration)
|
|
min_interval = float(args.min_interval)
|
|
hop_size = float(args.hop_size)
|
|
max_sil_kept = float(args.max_sil_kept)
|
|
num_worker = int(args.num_worker)
|
|
merge_short = bool(args.merge_short)
|
|
|
|
if __name__ == "__main__":
|
|
slice_audio_v2_(input_path, output_dir, num_worker, min_duration, max_duration, min_interval, threshold, hop_size, max_sil_kept,merge_short)
|