Through Google’s new and impressive MusicLM, AI can generate music pieces that are minutes long using just a simple prompt. Although the system is still facing many challenges, the sample songs that Google released are very promising. This AI can use prompts that are over a paragraph long, such as ones that describe the instruments and genre, but can create the same with a mere two-word prompt such as “melodic techno”. One of its challenges, however, is simulating human voices. The overall sound and the tone generally match a human voice well, but the quality is not quite the same. The lyrics may sound like English, but they don’t actually contain real sentences. Without concentrating on the words, the songs can sound very real.
Google was not the first to perform music generation with AI, and will definitely not be the last. Systems in the 90s were already composing pop songs and copying Bach with human-like capabilities. What makes MusicLM different is its high precision in adhering to the prompt and the quality of music that it produces. One of its unique features is allowing users to input a hummed tune and generate an electronic synth lead, string quartet, guitar solo, and more to go along with it.
Despite its amazing capabilities, Google announced that it has no current plans to release the system. The company faces potential issues of cultural appropriation and plagiarism, not wanting to take that risk just yet. On the bright side, researchers and developers of other music AI can definitely make use of Google’s system, with over 5,500 music text pairs being released publicly soon. Even though MusicLM may not be the future of musical AI, the potential for other systems to take that role is now tremendous. Although this system has yet to be fully released, it has the potential to massively impact the music industry. To get ahead of the curve with technological advancements, contact Aptimized below to start a conversation.