4 min readNew DelhiMay 14, 2023 09:35 AM IST
First published on: May 13, 2023 at 02:48 PM IST
Music is a universal language that we all enjoy and express in different ways. Whether we hum or sing along to our favourite tunes in the shower, on the road, or just for fun, we are participating in a musical experience. But what if we could also create music as easily as we vocalise it? Imagine being able to compose your own songs with the help of artificial intelligence.
MusicLM, the AI system that Google unveiled in January, helps you do just that. It not only takes text-based prompts like ChatGPT, but can also be conditioned by changing hummed or whistled melodies to match a text caption’s description of that style. Essentially, you can use it to convert a tune you’ve got on your mind into actual, high-fidelity music simply by humming it out.
Let’s take a look at how MusicLM works and how you can use it.
How does MusicLM work?
First detailed in an academic paper, MusicLM was trained on a dataset of 280,000 hours of music to generate ‘realistic’ songs for highly specific descriptions such as “a calming violin melody backed by a distorted guitar riff”. Users can specify instruments and genres like “classical” or “electronic” alongside the “vibe, mood, or emotion” they’re aiming for, to get the model to produce exactly what they have in mind.
In addition to that, Google also demonstrated its ability to transform whistled and hummed melodies according to the style described in a text caption. As if that wasn’t enough, the AI model could also be fed descriptions about art to generate music that fit the vibe of paintings.
However, despite the impressive abilities, there were some major concerns that arose as well. Google researchers acknowledged the many ethical challenges posed by models like MusicLM, including the fact that their creators could train them on copyrighted music to generate new songs. They also observed that the model had a tendency to occasionally generate music that was directly replicated from the songs it was trained on.
Taking these concerns into account, Google decided against releasing the model to the public, with the authors acknowledging the risks of potential “misappropriation of creative content.” However, that appears to have changed.
How to sign up for MusicLM?
While anyone can sign up to use MusicLM, you’ll be added to a waitlist first following registration. The AI model is available in the AI Test Kitchen app on the web, Android, and iOS. Below is how to get started:
1. Go to the official MusicLM page and click on the Get Started button.
2. If you see a pop-up saying that AI Test Kitchen is only available for limited testers, click on the “Register your interest” button.
3. Fill out a short survey about your location and motivation for joining the Test Kitchen.
4. Log in to your Google account at the end of the survey. If you don’t have one, create one first.
5. Wait for an email from Google that will grant you access to MusicLM. This may take time as the service is in high demand.
Congratulations, you have successfully joined the waitlist for MusicLM. Once you gain access, simply type your prompt into the text field and wait as the AI generates music for you. You will see two outputs for each prompt. Choose the one you prefer and award it a trophy to enhance the model.
While MusicLM may seem impressive on paper, its output reportedly can sometimes lack originality. This is probably due to a technical constraint. After all, the AI is only mixing music it has learned from and trying to produce something new, without really grasping what it is creating.
In this image created by Bing Image Creator, a robot listens with delight as a human performs a violin solo. (Express image) 
Share
