Google’s new AI tool can now generate music of any genre from text prompts and even transpose a tune into different instruments when it is hummed or whistled. According to Google Research, the recently introduced method, known as MusicLM, is a text-to-music creation system. The AI technology works by analysing the text and determining how lengthy and complex the composition is.
The study report states, “We introduce MusicLM, a model creating high-fidelity music from text descriptions such as a soothing violin tune supported by a distorted guitar riff.” It continued, “We demonstrate that MusicLM may be conditioned on both text and a melody in that it can modify whistled and hummed tunes in accordance with the manner defined in a text caption.”
In order to learn how to generate songs that make sense from word descriptions and to pick up subtleties like mood, melody, and instruments, the study report states that the new artificial intelligence has been trained on a dataset of 280,000 hours of music. It may be used for more than just making music videos. The ability of the algorithm to build on modern songs, whether they are hummed, sung, whistled, or played on an instrument, has been proven by Google researchers. Additionally, the study demonstrates that Google’s new AI tool is capable of taking a list of written instructions, such as “time to meditate,” “time to wake up,” “time to run,” and “time to give 100%,” and converting them into a melodic “story” or narrative that can endure for several minutes. Furthermore, Google’s new AI tool might be guided by a picture and caption combination or it might generate music that is “played” by a specific kind of instrument in a specific game.
It should be mentioned that Google is not the first company to do this. Initiatives like Google’s AudioLM, OpenAI’s Jukebox, or Riffusion—an AI that can make music by simply looking at it—have all made an attempt, according to TechCrunch. Due to technical limitations and a lack of training data, none have been able to create songs that are incredibly difficult to create in high-fidelity. Researchers believe that MusicLM may be the first to be able to do this.
ALSO READ: What’s next after Twitter removes the ‘Direct Message’ button on phones?
