fix dependecies

.gitignore

@@ -169,5 +169,4 @@ Info.plist
 /*.wav
 /*.flac
 output/
-*.exe
-fish_audio_preprocess
+*.exe

fish_audio_preprocess/cli/__main__.py

@@ -0,0 +1,45 @@
+import click
+import richuru
+from loguru import logger
+
+from fish_audio_preprocess.cli.merge_lab import merge_lab
+
+from .convert_to_wav import to_wav
+from .frequency import frequency
+from .length import length
+from .loudness_norm import loudness_norm
+from .merge_short import merge_short
+from .resample import resample
+from .separate_audio import separate
+from .slice_audio import slice_audio, slice_audio_v2, slice_audio_v3
+from .transcribe import transcribe
+
+
+@click.group()
+@click.option("--debug/--no-debug", default=False)
+def cli(debug: bool):
+    """An audio preprocessing CLI."""
+
+    if debug:
+        richuru.install()
+        logger.info("Debug mode is on")
+
+
+# Register subcommands
+cli.add_command(length)
+cli.add_command(frequency)
+
+cli.add_command(to_wav)
+cli.add_command(separate)
+cli.add_command(loudness_norm)
+cli.add_command(slice_audio)
+cli.add_command(slice_audio_v2)
+cli.add_command(slice_audio_v3)
+cli.add_command(resample)
+cli.add_command(transcribe)
+cli.add_command(merge_short)
+cli.add_command(merge_lab)
+
+
+if __name__ == "__main__":
+    cli()

fish_audio_preprocess/cli/convert_to_wav.py

@@ -0,0 +1,92 @@
+import subprocess as sp
+from pathlib import Path
+
+import click
+from loguru import logger
+from tqdm import tqdm
+
+from fish_audio_preprocess.utils.file import (
+    AUDIO_EXTENSIONS,
+    VIDEO_EXTENSIONS,
+    list_files,
+    make_dirs,
+)
+
+
+@click.command()
+@click.argument("input_dir", type=click.Path(exists=True, file_okay=False))
+@click.argument("output_dir", type=click.Path(exists=False, file_okay=False))
+@click.option("--recursive/--no-recursive", default=True, help="Search recursively")
+@click.option(
+    "--overwrite/--no-overwrite", default=False, help="Overwrite existing files"
+)
+@click.option(
+    "--clean/--no-clean", default=False, help="Clean output directory before processing"
+)
+@click.option(
+    "--segment",
+    help="Maximum segment length in seconds, use 0 to disable",
+    default=60 * 30,
+    show_default=True,
+)
+def to_wav(
+    input_dir: str,
+    output_dir: str,
+    recursive: bool,
+    overwrite: bool,
+    clean: bool,
+    segment: int,
+):
+    """Converts all audio and video files in input_dir to wav files in output_dir."""
+
+    input_dir, output_dir = Path(input_dir), Path(output_dir)
+
+    if input_dir == output_dir and clean:
+        logger.error("You are trying to clean the input directory, aborting")
+        return
+
+    make_dirs(output_dir, clean)
+
+    files = list_files(
+        input_dir, extensions=VIDEO_EXTENSIONS | AUDIO_EXTENSIONS, recursive=recursive
+    )
+    logger.info(f"Found {len(files)} files, converting to wav")
+
+    skipped = 0
+    for file in tqdm(files):
+        # Get relative path to input_dir
+        relative_path = file.relative_to(input_dir)
+        new_file = (
+            output_dir
+            / relative_path.parent
+            / relative_path.name.replace(
+                file.suffix, "_%04d.wav" if segment > 0 else ".wav"
+            )
+        )
+
+        if new_file.parent.exists() is False:
+            new_file.parent.mkdir(parents=True)
+
+        check_path = (
+            (new_file.parent / (new_file.name % 0)) if segment > 0 else new_file
+        )
+        if check_path.exists() and not overwrite:
+            skipped += 1
+            continue
+
+        command = ["ffmpeg", "-y", "-nostats", "-loglevel", "error", "-i", str(file)]
+
+        if segment > 0:
+            command.extend(["-f", "segment", "-segment_time", str(segment)])
+
+        command.append(str(new_file))
+
+        sp.check_call(command)
+
+    logger.info("Done!")
+    logger.info(f"Total: {len(files)}, Skipped: {skipped}")
+    logger.info(f"Output directory: {output_dir}")
+
+
+if __name__ == "__main__":
+    to_wav()

fish_audio_preprocess/cli/frequency.py

@@ -0,0 +1,124 @@
+import os
+from collections import Counter
+from concurrent.futures import ProcessPoolExecutor, as_completed
+from pathlib import Path
+from typing import Union
+
+import click
+import numpy as np
+from loguru import logger
+from tqdm import tqdm
+
+from fish_audio_preprocess.utils.file import list_files
+
+
+def count_notes_from_file(file: Union[Path, str]) -> Counter:
+    """Count the notes from a file
+    Args:
+        file (Union[Path, str]): The file to count the notes from
+    Returns:
+        Counter: A counter of the notes
+    """
+
+    import librosa
+    import parselmouth as pm
+
+    pitch_ac = pm.Sound(str(file)).to_pitch_ac(
+        voicing_threshold=0.6,
+        pitch_floor=40.0,
+        pitch_ceiling=1600.0,
+    )
+    f0 = pitch_ac.selected_array["frequency"]
+
+    counter = Counter()
+    for i in f0:
+        if np.isinf(i) or np.isnan(i) or i == 0:
+            continue
+
+        counter[librosa.hz_to_note(i)] += 1
+
+    return counter
+
+
+@click.command()
+@click.argument("input_dir", type=click.Path(exists=True, file_okay=False))
+@click.option("--recursive/--no-recursive", default=True, help="Search recursively")
+@click.option(
+    "--visualize/--no-visualize", default=True, help="Visualize the distribution"
+)
+@click.option(
+    "--num-workers",
+    default=os.cpu_count(),
+    help="Number of workers for parallel processing",
+)
+def frequency(
+    input_dir: str,
+    recursive: bool,
+    visualize: bool,
+    num_workers: int,
+):
+    """
+    Get the frequency of all audio files in a directory
+    """
+
+    import librosa
+    from matplotlib import pyplot as plt
+
+    input_dir = Path(input_dir)
+    files = list_files(input_dir, {".wav"}, recursive=recursive)
+    logger.info(f"Found {len(files)} files, calculating frequency")
+
+    counter = Counter()
+
+    with ProcessPoolExecutor(max_workers=num_workers) as executor:
+        tasks = [
+            executor.submit(count_notes_from_file, file)
+            for file in tqdm(files, desc="Preparing")
+        ]
+        for i in tqdm(as_completed(tasks), desc="Collecting infos", total=len(tasks)):
+            assert i.exception() is None, i.exception()
+            counter += i.result()
+
+    data = sorted(counter.items(), key=lambda kv: kv[1], reverse=True)
+
+    for note, count in data:
+        logger.info(f"{note}: {count}")
+
+    if not visualize:
+        return
+
+    x_axis_order = librosa.midi_to_note(list(range(300)))
+    x_axis_order = list(x_axis_order)
+    data = sorted(counter.items(), key=lambda kv: x_axis_order.index(kv[0]))
+
+    plt.rcParams["figure.figsize"] = [10, 4]
+    plt.rcParams["figure.autolayout"] = True
+    plt.bar([x[0] for x in data], [x[1] for x in data])
+    plt.xticks(rotation=90)
+    plt.title("Notes distribution")
+    plt.xlabel("Notes")
+    plt.ylabel("Count")
+
+    # Add grid to the plot
+    plt.grid(axis="y", alpha=0.75)
+    plt.grid(axis="x", alpha=0.75)
+
+    # Add percentage to the plot
+    total = sum(x[1] for x in data)
+    for i, v in enumerate(x[1] for x in data):
+        if v / total < 0.001:
+            continue
+
+        plt.text(
+            i - 1,
+            v + 1,
+            f"{v / total * 100:.2f}%",
+            color="black",
+            fontweight="bold",
+        )
+
+    plt.show()
+
+
+if __name__ == "__main__":
+    frequency()