This Neural Network Can Generate Lyrics Just Like Your Favourite Artiste

Artificial Intelligence is the notion of machines that exhibit intelligence and mimic cognitive functions that are usually associated with humans such as learning, reasoning, predicting, planning, recognising, and even problem-solving. Now, AI tools are being increasingly integrated into technological solutions and now it can even be induced in to generate lyrics that match the style of unique music artists.

Now, the researchers at the University of Waterloo, Canada, have recently achieved this feat by developing a  system that can generate such song lyrics.


A System That Can Generate Song Lyrics Of Your favourite Artist

  • The researchers explored how neural generative models could assist songwriters and musicians in writing song lyrics. The outputs of these models were used to generate song lyrics which were based on original artwork or compositions.
  • Instead of generating lyrics for an entire song, the model generated suggestions for lyrics which lines in the style of a specified artist. The system developed by Vechtomova and her colleagues were based on a neural network model called variational autoencoder (VAE) with artist embeddings, a multi-dimensional vector of real numbers and a CNN classifier which is trained to predict artists from MEL spectrograms of their song clips and can learn by reconstructing original lines of text.
  • The purpose of this model is that unusual and creative arrangements of words in the generated lines can inspire the songwriter to create original lyrics at a later stage.
  • The system also conditions, the generation on the style of a specific artist is done in order to maintain stylistic consistency of the suggestions.
  • The reason for using such generative models is mainly to augment the natural creative process when an artist gets inspired to write a song based on something they have read or heard.

According to Olga Vechtomova, University of Waterloo, one of the minds behind this innovation presses on the fact that the base of the research is the ultimate result  of curiosity that imbibes in each and every human being and that is to know whether a machine can generate lines that could sound like the lyrics of my favourite music artists.

While working on text generative models, The research team founded that neural networks had the capability to generate some of the most creative and impressive lines of text.

For the musically inclined research team, it was a natural next step to be able to explore whether a machine could learn the very essence of a specific music artist’s lyrical style, including choice of words, themes and sentence structure, to generate novel lyrics lines that sounded similar to the artist in question.

The motivation behind using artist embedding to condition the generation lyrics lines in the style of each artist was to reflect the differences between artist embeddings and their lyrical as well as musical styles.

  • Their findings suggested that artist embeddings were useful for generating lyrics that matched an artist’s style.
  • Many lines generated by the model were perfectly aligned with the artist as it was conditioned on in such a way that it reflected the themes generally addressed their music.

The very same system generated two poems which included the collection that was submitted to the NeurIPS 2018 Workshop on ML for Creativity and Design. Vechtomova created each of these poems by selecting lines generated by the VAE and arranged them in an artistically meaningful way was done to the individual lines were even not edited except for adding capitalization and punctuation marks.

According to Vechtomova, the generated lines often contains the words of an artist, these are used in an interesting new way, expressing novel thoughts not found in the original lyrics too. Some of the generated lines also convey new and powerful poetic imagery, expressed using stylistic devices such as metaphors and oxymorons, while remaining true to the style of the artist.


The system created by Vechtomova and team will be used by inspired artists who are composing lyrics for new songs. Rather than replacing lyric composers, the researchers hope that it will provide new ideas, which artists could modify and develop even more creative lyrics of their own.

According to Vechtomova, the system is not meant to replace a music artist, but to be used as a source of inspiration during the songwriting process. In the music world, this will be analogous to a synthesizer that will generate an infinite number of sounds, from which an artist can then create a song. Similarly, this tool will also generate an infinite number of novel lines that artists can use in any way they like to compose their own lyrics.

In the future, the teams plan to work on models that can learn new themes and vocabulary from additional sources and use them to generate lyrics in the style of a given artist. The team will also be exploring how such a system could potentially be used by music artists as a source of inspiration.

More Great AIM Stories

Martin F.R.
Martin F.R. works as a Technology Journalist at Analytics India Magazine. He usually likes to write detail-oriented articles which are well-researched in articulated formats. Other than covering updates on analytics, artificial intelligence & data science, his interests also include covering politics, economics, finance, consumer electronics, global affairs and issues regarding public policy matters. When not writing any articles, he usually delves into reading biographies of successful entrepreneurs or experiments with his new culinary ideas.

More Stories

Jayita Bhattacharyya
Hands-On Implementation Of Perceptron Algorithm in Python

Perceptron is the first neural network to be created. It was designed by Frank Rosenblatt in 1957. Perceptron is a single layer neural network. This is the only neural network without any hidden layer. Perceptron is used in supervised learning generally for binary classification.

3 Ways to Join our Community

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Telegram Channel

Discover special offers, top stories, upcoming events, and more.

Subscribe to our newsletter

Get the latest updates from AIM