lite-spleeter
: Lightweight musical instruments removing library
This is a lightweight Spleeter (a library that uses AI/ML to extract vocals from audio files) version that works with media files of any length, with very low and fixed RAM usage (around 1.5 GB RAM usage on my computer).
• Progress indicator
• Very lightweight on low-end Computers
• Smaller package and simpler
• No numpy errors
It processes 30-second segments sequentially. After processing all segments, it concatenates them. This approach of processing chunks instead of the whole audio file helps keep memory usage low.
Make sure you have ffmpeg installed.
Download package:
pip install lite-spleeter
lite-spleeter separate audio_example.mp3 -o audio_output_path
You can provide either a single or a list of files for batch processing
separate
command builds the model each time it is called, this process may be long, in this case If you have several files to separate, it is then recommended to perform all separation with a single call to separate:
lite-spleeter separate <path/to/audio1.mp3> <path/to/audio2.wav> <path/to/audio3.ogg> -o audio_output_path
To get help on the different options available with the separate command, type:
lite-spleeter separate --help
Read the original Spleeter repo for more info.