title: Extracting Individual VirtualDJ Stems
tags: audio

VirtualDJ can pre-process stems to save CPU power. Usefully
it stores the processed stems in a `.mp4` file with multiple
audio streams. We can extract these using `ffmpeg` and a simple
Python script:
```py
#!/usr/bin/env python
from subprocess import run, DEVNULL
import re
import sys
import os

what = """
0 vocals
1 hiperc
2 bass
3 synths
4 kick
""".strip().splitlines()
what = dict(x.split(" ") for x in what)

args = sys.argv[1:]
for arg in args:
  if not arg.endswith(".vdjstems"):
    arg += ".vdjstems"
  if not os.path.exists(arg):
    print(f"{arg} does not exist")
    continue

  basename = arg.split("/")[-1]
  fnstem = basename.split(".")[:-1]
  ext = fnstem.pop()
  fnstem = ".".join(fnstem)
  for k,v in what.items():
    ofn = f"{fnstem}.{v}.{ext}"
    cmd = ["ffmpeg","-n","-i",arg,"-c","copy","-map",f"0:{k}",ofn]
    run(cmd)
```