Google’s AI-powered text-to-music generator, MusicLM, is now available to users who sign up for its AI Test Kitchen; the product is available on the web, Android and iOS. Announced in January, MusicLM is one of the most anticipated generative AI products for music making.
According to the company’s announcement on Wednesday (May 10), users can generate convincing AI music with MusicLM using simple text prompts like “melodic techno,” “relaxing jazz” or even a more in-depth prompt like: “The main soundtrack of an arcade game. It is fast-paced and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember but with unexpected sounds, like cymbal crashes or drum rolls.”
MusicLM can also generate music based on a user humming a melody and adding a text prompt, directing MusicLM how they would like for the melody to be played — like as a guitar solo or an a cappella chorus, for instance. It can also create a song that corresponds to places, genres, epochs or even descriptions of other artworks like paintings.
MusicLM was trained on 5 million audio clips, equal to 280,000 hours of listening, plus an additional dataset of 5,500 music clips with captions written by professional musicians. Each caption describes the music in sentences and includes aspects like the music’s genre, mood and rhythm.
MusicLM’s quick release may come as a surprise to some who read through its academic paper published in January, in which the team noted that it “acknowledges the severe risks associated with our model,” including “music generation for cultures underrepresented in the training data” and “cultural appropriation,” and said that they had “no plans to release models at this point.”
The January paper also nodded at the team’s possible future plans to improve the model, citing that it “may focus on lyrics generation, along with improvement of text conditioning and vocal quality.”
Generative AI, a form of emerging artificial intelligence technology that can create images, answers, code, music and more at the touch of a button, has become a hotly debated topic in the music industry in recent months. Some of the largest music companies like Universal Music Group, Warner Music Group and others have addressed the potential pitfalls and opportunities of the technology in recent weeks, and over 80 entertainment organizations — like the RIAA, NMPA, SESAC, BMI, ASCAP, UMG and more — have signed on to a list of seven principles called the Human Artistry CampAIgn.
At the core of the debate is whether or not generative AI, including MusicLM and its competitors, could automate away jobs for musicians and weaken rights holders’ ability to protect and monetize their copyrights and rights of publicity.
