VocalX Instrument Separator
A command-line audio exorcist that pulls vocals, drums and instruments apart into separate stems.
What VocalX Instrument Separator is
The command-line engine that became the split feature inside Cypher Music. Point it at a song and Deezer’s Spleeter model separates the mix into two, four or five stems, then builds a combined instruments track and a background track from what is left. It runs on the CPU, pulls the audio out of a video first when you hand it one, and narrates what it is doing instead of sitting silent.
VocalX Instrument Separator features
What it actually does.
Two, four or five stems
Vocals and accompaniment; or vocals, drums, bass and other; or all of that with the piano split out as well.
Instruments and background
The non-vocal stems are mixed into one Instruments track, and a Background track is worked out by subtracting vocals and instruments from the original.
RAM guard
Checks the memory the machine actually has and recommends a stem count it can finish, rather than letting a five-stem run fall over halfway.
Video to audio
MP4, MKV, AVI, MOV, WMV, FLV, WebM and M4V in; MP3, WAV or FLAC out, at the bitrate you choose.
Models fetched once
The pretrained Spleeter models download on first use and are reused on every run after that.
Two ways to drive it
A guided menu with tab completion on every path, or a direct command for scripts: cli.py separate song.mp3 --stems 5.
How to use VocalX Instrument Separator
From a fresh install to the thing working.
1 Install it
- Install Python 3.8 or newer, and FFmpeg on your PATH.
- Clone the repository and run python setup.py. It offers a local venv or a conda environment called vocalx, checks for FFmpeg and installs the requirements.
- Prefer to do it by hand? Create a virtual environment and pip install -r requirements.txt.
2 Separate a song
- Run python run.py for the guided menu.
- Give it the path to an MP3, WAV, FLAC or OGG file. Tab completion works on every path prompt.
- Choose 2, 4 or 5 stems. The RAM guard tells you what this machine can handle.
- The stems, plus the combined Instruments and Background tracks, are written out when it finishes.
3 Script it
- python cli.py separate "path/to/song.mp3" --stems 5 runs a separation with no menu.
- Use --stems 2 when you only need vocals and accompaniment. It is the fastest.
4 Pull the audio out of a video
- Choose video to audio from the interactive menu.
- Point it at the video, then pick MP3, WAV or FLAC and a bitrate.
- Separate the result like any other song.
Built with
Licence
Made by Kencypher. Not open source.
VocalX Instrument Separator was written by Kencypher. The source is not published and is not licensed for reuse. You have permission to run it — that is the whole of it.
You may
- Run it on as many of your own machines as you like
- Use it for anything, including commercial work
- Pass an unmodified copy to a friend, free
You may not, without asking
- Sell it or charge for access
- Re-upload or mirror it anywhere
- Rebrand it or publish it as your own
- Decompile it or lift its assets
This is enforced. Re-uploading it, repackaging it, stripping the author's name off it or passing it off as your own is copyright infringement — protected without registration under the Berne Convention — and is pursued with DMCA takedowns to the host, CDN, app store and search engine, platform reports for impersonation, and legal proceedings where the infringement is commercial, repeated, or involves falsely claiming authorship.
Asking first is free, and the answer is usually yes — kencypher56@gmail.com. The only official source is this store; a copy from anywhere else is not one Kencypher published.
Good to know
- This is the desktop engine behind the split feature in Cypher Music, which now runs the same Spleeter model on the phone and the desktop without a terminal.
- Everything runs on your own CPU. Once the models have downloaded, nothing needs the network.
- The Background track is arithmetic, not a model: whatever is left once the vocals and instruments are taken out of the original.
- The models land in pretrained_models/ on the first run, a few hundred megabytes once. Delete that folder and the next run fetches them again.
- It reads MP3, WAV, FLAC and OGG. Anything else is converted to audio first, which is what the video-to-audio step is for.
- The terminal narrates each stage rather than printing a silent bar, so a five-stem run on a slow machine still shows which part it is on.
- Spleeter and FFmpeg belong to their own authors and keep their own licences.