SoundPicker

About

SoundPicker is designed to work with audio files and use already trained state of the art models for performing separations and variouse sound extractions. The current version tests Spleeter a source separation library with pretrained models written in Python and uses Tensorflow. This implementation includes Vocals / accompaniment separation API usage.

Running

Launching ZeroMQ Message Streamer

python3 streamer_device.py

Launching Worker (ZeroMQ Consumer)

To launch workers the below command needs to be executes one or multiple times:

python3 worker.py

Launching REST API (ZeroMQ Publisher)

gunicorn --timeout 120 --bind 127.0.0.1:8080 wsgi:app

Launching Nginx

nginx -c $PWD/nginx.conf

Usage

Upload Source File to Server

Sample request:

curl -X POST -F source_file=@/tmp/myfile.mp3 http://127.0.0.1:8888/upload

Sample response:

{
    "source_id": "3e50836e-7b65-4fac-9e3e-81f7f5bb696b"
}

Invoke ML Model

Sample request:

curl -i -H "Content-Type: application/json" -X POST \
  -d '{"source_id": "3e50836e-7b65-4fac-9e3e-81f7f5bb696b", "split_type": "spleeter:4stems"}' \
  http://127.0.0.1:8888/invoke

Sample response:

{
    "request_id":"2d85c022-bdee-4462-ab85-f9a12b6402d6"
}

Check Request Status

Sample request:

curl -i http://127.0.0.1:8888/status/2d85c022-bdee-4462-ab85-f9a12b6402d6

Sample response:

{
    "output_artifacts":[
        "http://127.0.0.1:8888/artifacts/2d85c022-bdee-4462-ab85-f9a12b6402d6/myfile/accompaniment.wav",
        "http://127.0.0.1:8888/artifacts/2d85c022-bdee-4462-ab85-f9a12b6402d6/myfile/vocals.wav"
    ],
    "request_id":"2d85c022-bdee-4462-ab85-f9a12b6402d6",
    "status":"success"
}

Download Results

wget http://127.0.0.1:8888/artifacts/2d85c022-bdee-4462-ab85-f9a12b6402d6/myfile/accompaniment.wav