A simple cli that will run spleeter without too much hassle

spleeter-bash.md

Spleeter-Bash

Simply run spleeter-bash.sh "/path/to/file" "/path/to/file/x/if/multiple/", select the model you'd like to use, and done! The script will make sure the pretrained_models are saved with spleeter-bash.sh (so they don't need to be redownloaded every run) and will make sure that the output gets set in a subdirectory of the source file. (In case of multiple files, the first files' directory will be used, since Spleeter only supports one output directory)

Prerequisites

• A conda installation with spleeter installed
• Possibly "conda init" run once or conda added to path, in case it doesn't work (f.e. gives a permissionerror)

spleeter-bash.sh

# Select model
model_selected=false
while [ $model_selected = false ]; do
    echo "The following models are available:"
    echo "[1] - vocals / accompaniment (2 stems)"
    echo "[2] - vocals / bass / drums / other (4 stems)"
    echo "[3] - vocals / bass / drums / piano / other (5 stems)"
    read -p "Select a model: " model
    if [ $model = "1" ] || [ $model = "[1]" ]; then
        model_selected=true
        model="spleeter:2stems"
        output_dirname="2stems"
    elif  [ $model = "2" ] || [ $model = "[2]" ]; then
        model_selected=true
        model="spleeter:4stems"
        output_dirname="4stems"
    elif [ $model = "3" ] || [ $model = "[3]" ]; then
        model_selected=true
        model="spleeter:5stems"
        output_dirname="5stems"
    fi 
done

# Activate conda env
source ~/anaconda3/etc/profile.d/conda.sh
conda activate base

# Aggregate all files to be processed
files=()
for file in "$@"
do
    files+=("$file")
done
  
cd "$(dirname $0)"

# Use first arguments dir for output
output_dir="$(dirname "$1")/output-$output_dirname"

py -3.7 -m spleeter separate -i "${files[@]}" -o "$output_dir" -p $model