Are you looking to turn whatever is in your mind into a music melody, with Google’s MusicLM, you can create songs by providing written prompts that describe instruments, genres, emotions, and other relevant details and this has been introduced recently.
Music is a language that transcends borders and unites people across the world. We all have our unique ways of expressing and enjoying music, whether it’s humming in the shower or singing along or while driving. However, what if we could also create music effortlessly like we express it? With the help of artificial intelligence, we can now compose our own songs easily.
Google’s MusicLM, an AI system allows us to do just that. It not only responds to text-based prompts like ChatGPT but it can also be conditioned by converting hummed or whistled melodies to match a text caption’s description of that style. Essentially, by humming a tune that’s in your mind, you can use MusicLM to turn it into high-quality music that matches your desired style.
MusicLM operates by using a vast dataset of 280,000 hours of music to produce “realistic” songs based on specific descriptions, such as “a calming violin melody backed by a distorted guitar riff.” Users can input details about instruments, genres, and emotions they want to convey to get the model to create their desired sound of their own.
Furthermore, MusicLM can also convert hummed or whistled melodies into music that matches the style described in a text caption. Additionally, Google demonstrated its capability to generate music based on descriptions of artwork, aligning with the painting’s vibe. It can turn your creative thinking into music.
Despite its impressive capabilities, there are also concerns surrounding MusicLM. Google researchers highlighted the ethical challenges posed by models such as MusicLM, including the possibility that they could be trained on copyrighted music to create new songs. They also noted that the model occasionally generated music that was similar to the songs it was trained on.
It seems that Google has had a change of heart regarding the public release of MusicLM, despite the ethical concerns. The model is now available as part of the open-source TensorFlow platform, with the hope that researchers and developers will use it to advance the field of AI-generated music.
To address the copyright issue, the model was trained using only non-copyrighted music, and Google has emphasized that any use of the model for copyrighted content would be illegal and against their terms of service.
To sign up for MusicLM, you can register your interest on the official MusicLM page and then complete a short survey about your location and motivation for joining the Test Kitchen. After logging in to your Google account (or creating one if you don’t have one), you’ll be added to a waitlist to gain access to MusicLM. This process may take some time due to high demand.
Once you receive an email from Google granting you access to MusicLM, you can use the AI Test Kitchen app on the web, Android, or iOS to generate music based on your prompts. Simply input your desired details into the text field and wait for the AI to create two outputs for each prompt. Then, select your preferred output and award it a trophy to improve the model.
MusicLM: Google's AI tool that can generate music from text prompts.
— Saadh Jawwadh (@SaadhJawwadh) May 11, 2023
Initially, #Google released a research paper and proof of concept on this and now it is testing an interactive web app. pic.twitter.com/c94If31bSP
Link to Subscribe now – https://blog.google/technology/ai/musiclm-google-ai-test-kitchen/.