Google has invented a Robot Music Composer

Google has invented a Robot Music Composer

Composing music is a demanding job, but this might change now. Google has invented a robot music composer to automatically compose music in any genre of your choice, such as jazz, rock, heavy metal, folk, hip-hop and many more. You just need to input a poem to Google’s robot music composer and select a genre of your choice. Google’s robot music composer will use a trained neural network to automatically generate a music composition for you. Yes, it’s that simple. So, now you can explore your inner Beethoven!


*This GIF is shown for illustration purpose only. This is not an actual robot from Google.

Alex Fletcher (Hugh Grant), in the American romantic comedy Music and Lyrics, struggled to compose music for a song for Cora Corman. Writing lyrics and composing music for a song involve totally different skillsets. A song may be written in a span of a few hours to a few days. However, composing music involves more time and effort. Often composing music involves using multiple instruments like a piano, keyboard, a drum kit, a guitar, a violin and many more just to get the perfect sound with a perfect loudness, pitch, rhythm, or tempo. For instance, as per a BBC Music report, well-known singer Charlie Puth wrote the famous hit song “See you again” in just 10 minutes. However, composing music for the song would have taken them a few weeks at the least. Google’s music composer would make composing music easier. Google just filed a patent application on this software-based robot music composer.


Google’s robot music composer consists of a music generation unit (refer figure below) that takes textual data as an input and provides music as an output. Any sort of text can be input to the music generation unit. For instance, the textual data may be a poem, a text conversation from a messaging application, or even news articles from a web browser. The textual data is then analyzed by a feature extractor unit and a neural network for composing music for the user.

The feature extractor unit, as its name suggests, extracts structural features of the input textual data. The structural features include patterns and sequences of words (such as rhyming words) present in the text. Then, the neural network compares these structural features of the text with lyrics of known songs that are stored in a database. The database consists of a huge number of known songs with a detailed description of music features (such as tempo, pitch, loudness etc.) associated with each song in an organized manner. The neural network has an ability to learn and predict such music features for the text input by the user. In fact, the Google’s robot music composer consists of a number of neural networks and each neural network is an expert in a specific genre of music. For instance, if the user wants to compose a music for a poem in a hip-hop genre then the Google’s robot music composer will select a neural network that is expert in the hip-hop genre.


So what do you think about this technology from Google? Let us know in the comment section.

Patent Information
Publication Number: US20180190249
Patent Title: Machine Learning to Generate Music from Text
Publication date: 2018-07-05
Filing date: 2016-12-30
Inventors: Dominik Roblek; Douglas Eck
Original Assignee: Google Inc.