sound-spleeter.md

Sound Spleeter

This service uses Spleeter to perform source separation on audio files.

It is part of our third party DNN Model Services.

Welcome

The service receives an audio sample as input for source separation pre-trained model.

What’s the point?

The service separates the vocals and the accompaniment from the input audio sample using machine learning techniques.

The service outputs two audio files (gRPC bytes) with both separations.

How does it work?

The user must provide the following inputs in order to start the service and get a response:

Inputs:

• audio_url or audio: An URL with an audio file (mp3) or an audio bytes array.

You can use this service from SingularityNET DApp, clicking on SNET/SoundSpleeter.

You can also call the service from SingularityNET CLI (snet).

Assuming that you have an open channel to this service:

$ snet client call --save-field vocals vocals.wav snet sound-spleeter default_group spleeter '{"audio_url": "https://bh.singularitynet:7000/Resources/audio_example.mp3"}'
Price for this call will be 0.00000001 AGI (use -y to remove this warning). Proceed? (y/n): y

The WAV audio file will be saved in vocals.wav!

What to expect from this service?

Inputs:

• audio: Slow Motion Dream by Steven M Bryant

Response:

• audio bytes array: Vocals
• audio bytes array: Accompaniment