Waterloo Researchers Create LyricJam, an AI-Supported Tool for Writing Song Lyrics

University of Waterloo researchers developed a technology WITH A real-time system of providing inspiration to music artists when in the process of songwriting. Called LyricJam, the program utilizes artificial intelligence (AI) in producing lyrics for live instrumental music. The development of LyricJam was led by a Waterloo Engineering professor and members of the university’s Natural Language Processing Lab.

The lab researchers have been studying artificial intelligence for years in their application for creative works. The research led to the invention of a system that observes artists’ style and musical expressions, whilst having the capability to creates lyrics based on AI observations.

The lyric-producing system relies on different aspects of music, such as tempo, instrumentation, and chord progression. The lyrics are then synthesized as a means of mirroring emotions and mood conveyed by the live music.

Thereafter, Professor Olga Vechtomova, of the Waterloo Engineering Dept. with a cross-appointment in Computer Science, along with Waterloo graduate students Dhruv Kumar and Gaurav Sahu were able to develop LyricJam.

How Does LyricJam Work?

The system constantly receives raw audio clips while the artist/s play their sample of instrumental music. The AI’s neural network processes all musical information in creating lyric lines. An artists can then utilize the lyrics provided by LyricJam or simply take inspiration from it.

The AI’s neural network is designed to learn the words, stylistic devices, and lyrical themes that are correlated to various details of music.

How Professional Musicians are
Responding to LyricJam

During a user study held by the research team, participants were inspired to improvise on the lyrics generated by by LyricJam. Some professional artists also changed a few chords and took a new creative direction than what was originally planned. Some others based their lyrics on the AI-suggested lines using improvisations that really went on the same direction suggested by the AI.

The overall view gathered from musicians is that the co-creative aspect of the experience, made them perceive LyricJam as an uncritical jamming partner. As such, they were encouraged to continue playing the instruments even if they were not into writing lyrics.

More than 1,500 users have tried LyricJam since it was released and made accessible in June, at https://lyricjam.ai . Moreover, the Waterloo Research will be presented next month at the International Conference on Computations Creativity.