my_spleeter.py

import os
import re
import sys
import glob
import shutil
import subprocess
from pathlib import Path
import time
import logging
from tqdm import tqdm
import subprocess

from pydub import AudioSegment

import tensorflow as tf
tf.get_logger().setLevel('ERROR')

################################################################################def
now_dir = os.getcwd()
sys.path.append(now_dir)
#os.chdir(f'{now_dir}/spleeter')

# turn off tensor warning
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'


################################################################################def

# valid_chars includes: </>, <space>, <_>, <.>
valid_chars = '/abcdđefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789 _aàáảãạăằắẳẵặâầấẩẫậeèéẻẽẹêềếểễệiìíỉĩịoòóỏõọôồốổỗộơờớởỡợuùúủũụưừứửữựyỳýỷỹỵ'

def rename_file(file_path,valid_chars):
    """
    valid_chars includes: </>, <space>, <_>, <.>
    valid_chars = '/abcdđefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789 _aàáảãạăằắẳẵặâầấẩẫậeèéẻẽẹêềếểễệiìíỉĩịoòóỏõọôồốổỗộơờớởỡợuùúủũụưừứửữựyỳýỷỹỵ'

    """
    try:
        name, ext = os.path.splitext(file_path)

        # Duyệt qua từng ký tự trong tên file và chỉ giữ lại các ký tự hợp lệ
        new_name = ''.join([char if char.lower() in valid_chars else '' for char in name])    
        # Thay khoảng trắng bằng '_'
        new_name = new_name.replace(' ', '_')
        # Kết hợp lại tên mới với phần mở rộng không thay đổi
        new_file_path = new_name + ext
        # Đổi tên file
        os.rename(file_path, new_file_path)
    
    except Exception as e:
        print(f"\nError processing {file_path}: {e}")



def mp_rename(in_dir,valid_chars):
    """
    valid_chars includes: </>, <space>, <_>, <.>
    valid_chars = '/abcdđefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789 _aàáảãạăằắẳẵặâầấẩẫậeèéẻẽẹêềếểễệiìíỉĩịoòóỏõọôồốổỗộơờớởỡợuùúủũụưừứửữựyỳýỷỹỵ'

    """
    import os
    import glob

    from tqdm import tqdm
    from multiprocessing import Pool
    
    #os.makedirs(out_dir, exist_ok=True)
    files = glob.glob(f'{in_dir}/*.*')

    print('Start multiprocessing')
    
    with Pool() as pool:
        results = [pool.apply_async(rename_file, args=(file, valid_chars)) for file in files]
    
        # Dùng tqdm để hiển thị tiến trình
        for result in tqdm(results, desc="Processing files"):
            result.get()  # Đảm bảo chờ quá trình hoàn thành


################################################################################def

def set_cache(cache_dir):
  """
  1. Backup and restore your checkpoints
  2. Change cache default: /root/.cache/torch/hub
  set_cache('/content') -> '/content/hub'

  """
  import os
  import torch
  print("Before: ",torch.hub.get_dir())
  os.makedirs(cache_dir, exist_ok=True)
  os.environ['TORCH_HOME'] = cache_dir
  #torch.hub.set_dir(cache_dir)  
  print("After: ",torch.hub.get_dir())

################################################################################def

def wav2mp3(input_file, output_file):
  """ 256kbps, os.devnull """

  with open(os.devnull, 'w') as devnull:    
    subprocess.run(['ffmpeg', '-y', '-i', input_file, "-b:a", "256k", '-ac', '1', '-ar', '44100', output_file],stdout=devnull, stderr=devnull, check=True)


################################################################################def
## disable logging from spleeter/spleeter/audio/ffmpeg.py
import logging as logging
logger = logging.getLogger(__name__)
logging.disable(logging.INFO)  

from spleeter.separator import Separator
#separator = Separator('spleeter:2stems')

stems2 = Separator('spleeter:2stems')
stems4 = Separator('spleeter:4stems')
stems5 = Separator('spleeter:5stems')


def get_vocal(separator,song_file,out_dir):
  """  
  spleeter_file('/content/test/wav/song_2173577437.wav','acapella')
  max_time = 10 minutes? longer will OOM
  
  """
  # https://github.com/deezer/spleeter/wiki/4.-API-Reference
 
  os.makedirs(out_dir, exist_ok=True)
  
  filename = os.path.basename(song_file)
  output_file = f'{out_dir}/{filename[:-4]}.mp3'

  os.system(f"cp '{song_file}' current.mp3")  
  separator.separate_to_file('current.mp3', 'temp')    
  vocal_file = f'temp/current/vocals.wav' # clean vocal

  wav2mp3(vocal_file, output_file)
  #print(f'Output file: {output_file}')



def spleeter_folder(src_dir,dst_dir):
  """
  """
  import glob
  from tqdm import tqdm

  import logging as logging

  logger = logging.getLogger(__name__)
  logging.disable(logging.INFO)

  # pretrained model: spleeter/pretrained_models/2stems

  # Extract vocal
  print('Note: this fixed the wrong file content.')
  mp3 = glob.glob(f'{src_dir}/*.mp3')
  wav = glob.glob(f'{src_dir}/*.wav')
  files = mp3 + wav
  files.sort()
  
  for file in tqdm(files):
    #print(f'music file:{file}')
    get_vocal(stems2,file,dst_dir)  
  

################################################################################def
def get_stems(separator,song_file,out_dir):
  """    
  max_time = 10 minutes? longer will OOM
  get_stems(stems2,'/content/mixed.mp3','/content/demo')
  get_stems(stems4,'/content/mixed.mp3','/content/demo')
  get_stems(stems5,'/content/mixed.mp3','/content/demo')
  
  """
  # https://github.com/deezer/spleeter/wiki/4.-API-Reference 

  #import shutil
  #import glob

  os.makedirs(out_dir, exist_ok=True)
  
  file_name = os.path.basename(song_file)
  input_name = os.path.splitext(file_name)[0]

  temp_dir = 'temp/current'
  if os.path.exists(temp_dir) and os.path.isdir(temp_dir):
    shutil.rmtree(temp_dir)
  os.system(f"cp '{song_file}' current.mp3")  
  separator.separate_to_file('current.mp3', 'temp')    
  files = glob.glob('temp/current/*.*')
  
  for stem_file in files:
    file_name = os.path.basename(stem_file)
    stem_name = os.path.splitext(file_name)[0]
    output_file = f"{out_dir}/{input_name}_{stem_name}.mp3"
    wav2mp3(stem_file, output_file)


def extract_accompaniment(separator,src_dir,dst_dir):
  """
  """
  import glob
  from tqdm import tqdm

  import logging as logging

  logger = logging.getLogger(__name__)
  logging.disable(logging.INFO)

  # Extract all stems
  print('Note: this fixed the wrong file content.')
  mp3 = glob.glob(f"{src_dir}/*.mp3")
  wav = glob.glob(f"{src_dir}/*.wav")
  files = mp3 + wav
  files.sort()
  
  for file in tqdm(files):
    get_stems(separator,file,dst_dir)
################################################################################def

def extract_stem(separator,song_file,out_dir,keep_name='vocals'):
  """    
  max_time = 10 minutes? longer will OOM
  extract_stem(stems2,'/content/mixed.mp3','/content/demo',keep_name='vocals')
  extract_stem(stems4,'/content/mixed.mp3','/content/demo',keep_name='vocals')
  extract_stem(stems5,'/content/mixed.mp3','/content/demo',keep_name='other')
  keep_name = 'bass'|'drums'|'piano'|'vocals'|'other'
  
  """
  # https://github.com/deezer/spleeter/wiki/4.-API-Reference 

  #import shutil
  #import glob

  if keep_name not in ['bass','drums','piano','vocals','other']:
    raise ValueError("stem name must be: 'bass', 'drums', 'piano', 'vocals', or 'other'")
  
  os.makedirs(out_dir, exist_ok=True)
  
  file_name = os.path.basename(song_file)
  input_name = os.path.splitext(file_name)[0]

  temp_dir = 'temp/current'
  if os.path.exists(temp_dir) and os.path.isdir(temp_dir):
    shutil.rmtree(temp_dir)
  os.system(f"cp '{song_file}' current.mp3")  
  separator.separate_to_file('current.mp3', 'temp')    
  files = glob.glob('temp/current/*.*')
  
  for stem_file in files:
    file_name = os.path.basename(stem_file)
    stem_name = os.path.splitext(file_name)[0]
    if stem_name == keep_name:
        output_file = f"{out_dir}/{input_name}_{stem_name}.mp3"
        wav2mp3(stem_file, output_file)


def extract_one(separator,src_dir,dst_dir,keep_name='piano'):
  """
  keep_name = 'bass'|'drums'|'piano'|'vocals'|'other'
  
  """
  import glob
  from tqdm import tqdm

  import logging as logging

  logger = logging.getLogger(__name__)
  logging.disable(logging.INFO)

  if keep_name not in ['bass','drums','piano','vocals','other']:
    raise ValueError("stem name must be: 'bass', 'drums', 'piano', 'vocals', or 'other'")

  # Extract all stems
  print('Note: this fixed the wrong file content.')
  mp3 = glob.glob(f"{src_dir}/*.mp3")
  wav = glob.glob(f"{src_dir}/*.wav")
  files = mp3 + wav
  files.sort()
  
  for file in tqdm(files):
    extract_stem(separator,file,dst_dir,keep_name)


################################################################################def
def saxophone(separator,song_file,out_dir):
  """    
  stems = 'bass'|'drums'|'piano'|'vocals'|'other'
  saxophone = other + vocals
  
  """
  # https://github.com/deezer/spleeter/wiki/4.-API-Reference 

  #import shutil
  #import glob

  
  os.makedirs(out_dir, exist_ok=True)
  
  file_name = os.path.basename(song_file)
  input_name = os.path.splitext(file_name)[0]

  temp_dir = 'temp/current'
  if os.path.exists(temp_dir) and os.path.isdir(temp_dir):
    shutil.rmtree(temp_dir)
  os.system(f"cp '{song_file}' current.mp3")  
  separator.separate_to_file('current.mp3', 'temp')    
  files = glob.glob('temp/current/*.*')

  # Load audio stems
  other = AudioSegment.from_file("temp/current/other.wav")
  vocals = AudioSegment.from_file("temp/current/vocals.wav")
  
  remix = other.overlay(vocals)
  remix = remix.set_frame_rate(44100).set_channels(1)
  remix.export(f'{out_dir}/{input_name}.mp3','MP3')


def extract_saxophone(separator,src_dir,dst_dir):
  """
  keep_name = 'bass'|'drums'|'piano'|'vocals'|'other'
  extract_saxophone(stems5,src_dir,dst_dir)
  
  """
  import glob
  from tqdm import tqdm

  import logging as logging

  logger = logging.getLogger(__name__)
  logging.disable(logging.INFO)

  # Extract all stems
  #print('Note: this fixed the wrong file content.')
  mp3 = glob.glob(f"{src_dir}/*.mp3")
  wav = glob.glob(f"{src_dir}/*.wav")
  files = mp3 + wav
  files.sort()
  
  for file in tqdm(files):
    saxophone(separator,file,dst_dir)

################################################################################def

def combine_audio(input_file1, input_file2, output_file):
    """
    """
    ffmpeg_command = [
        'ffmpeg',
        '-y',   # overwrite output
        '-i', input_file1,
        '-i', input_file2,
        '-filter_complex', '[0:0][1:0]amerge=inputs=2',   # combine two audio stream
        '-ac', '1',  # Set the output audio channels to 1 (mono)
        '-ar', '44100',  # Set the output audio sample rate to 44100 Hz
        '-codec:a', 'libmp3lame',  # Set the output audio codec to libmp3lame (MP3)
        '-vol', '256',  # range volume 0-256
        '-loglevel', 'debug',
        output_file
    ]
    
    result = subprocess.run(ffmpeg_command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
    
    if result.returncode != 0:
        print(f"FFmpeg error: {result.stderr.decode()}")
        return
		
################################################################################def