Google creates an AI-based tool to generate music from text inputs 

Tech major Google always tries to be one or more steps ahead of others when it comes to introducing cutting-edge innovations. Following the success of their AI-based language moderator and translator late last year, Google is now back with another state-of-the-art facility driven by artificial intelligence. It is a AI-based tool that can produce minutes-long musical pieces from text prompts. In fact, it can also transform a whistled or hummed melody into instrumental pieces, something like DALL-E does by creating images from written prompts. How cool is that!  

Google had named this tool MusicLM. While it is not commercially available worldwide yet, Google has shared a glimpse of what it might feel like by uploading a series of samples on their official website. These include 30 seconds snippets of music derived out of paragraph long descriptions that include a specified genre, vibe, and required instruments. Five minute long pieces have also been produced from only a couple of words such as “melodic techno.” However, what has gained most popularity is the demo of a story mode where the tool morphs between prompts as per a given script. The brand-new AI-driven music model of Google is also equipped to imitate 10 second long sounds of instruments as per the simple mention of their names, such as cello or guitar. Not only that, google also unveiled samples such as 8 seconds clip of a background noise that would suit a prison escape scene to how a piano would sound like when played by a beginner as opposed to a professional. Wait, that isn’t all either. MusicLM can also interpret phrases like “futuristic club” and “accordion death metal” and produce authentic music accordingly. The fascinating music model can also replicate human vocals that although sounds quite accurate from the tonal or melodic point of views have their own distinct attribute – something that experts describe as grainy or static.  

Now, if you get a chance to hear some of them, you may find the lyrics gibberish. But it is more like exotic; something that happens when you hear a song in a language completely unknown to you, but you enjoy nonetheless.  

For those unaware, AI-generated music is not new and has been around for quite sometime now. Previously, it has created pop songs, imitated Bach’s music better than people, and even accompanied during live performances. Then there is also tools like StableDiffusion that are so advanced that uses AI-image generation engine to turn texts into spectrograms that are in turn converted into music. But it looks like Google’s MusicLM are going to surpass all its precursors as it is multi-functional and also performs better in terms of quality and adherence to the texts.  

Interestingly, Google has no plans to release it into the market for public use in the near future and is being extra careful to avoid “potential misappropriation of creative content” or what we know as plagiarism. It also wants to avoid any kinds of musical misrepresentations.