demucs
Separate audio into stems (vocals, background, drums, bass, etc.)
using fal.ai's Demucs model. Requires a fal.ai API key.
By default extracts vocals and background (other) stems.
Usage
1egaki demucs [audio]
Arguments
| Argument | Required | Description |
[audio] | No | audio |
Options
| Option | Default | Description |
--stems [list] | - | Comma-separated stems to extract: vocals, drums, bass, other, guitar, piano. Default: vocals,other |
--model [model] | - | Demucs model variant. Default: htdemucs_6s. Options: htdemucs, htdemucs_ft, htdemucs_6s, hdemucs_mmi, mdx, mdx_extra, mdx_q, mdx_extra_q |
--output-format [format] | mp3 | Output audio format: mp3 or wav |
-o, --output [dir] | - | Output directory for separated stems. Default: current directory |
--shifts [n] | - | Number of random shifts for quality (higher = better but slower). Default: 1 |
--overlap [n] | - | Segment overlap 0.0-1.0 (higher = better quality). Default: 0.25 |
--stdin | - | Read audio from stdin instead of a file path |
--json | - | Output result metadata as JSON to stdout |
Global Options
| Option | Default | Description |
-h, --help | - | Display this message |
-v, --version | - | Display version number |
Examples
1# Separate vocals and background from an audio file
1egaki demucs song.mp3
1# Extract all stems (vocals, drums, bass, guitar, piano, other)
1egaki demucs song.mp3 --stems vocals,drums,bass,other,guitar,piano
1# Extract only vocals as WAV
1egaki demucs song.mp3 --stems vocals --output-format wav
1# Save to a specific directory
1egaki demucs song.mp3 -o stems/
1# Higher quality (slower)
1egaki demucs song.mp3 --shifts 5