Hum2Song: Multi-track Polyphonic Music Generation from Voice Melody Transcription with Neural Networks

Hum2Song! is an AI-powered web application that is able to compose the musical accompaniment of a melody produced by a human voice.

Demo: https://www.carlostoxtli.com/hum2song/

Blog post: https://medium.com/@carlostoxtli/hum2song-multi-track-polyphonic-music-generation-from-voice-melody-transcription-with-neural-7a777445550b

Instructions

Once the site has loaded, you only need to click on the "Record Audio" button, start humming/singing and click on the "Stop" button to start the conversion process. Google Chrome browser is recommended. This is a client-side only application, this means that all the files are stored locally and the processing happens in your browser. The first time you load this website, it downloads ~100Mb in models and sounds, that takes in average ~1 minute depending on your internet connection. Please be patient.

Installation

This is a basic static HTML application, so you only need to place the files in your web server and load it. If you don't have a web server enabled in your computer you can easily run this command within the folder.

In python 2

sudo python -m SimpleHTTPServer 80

In python 3

sudo python -m http.server 80

And then navigate to you http://localhost/

Screenshot: