
1.WHAT IS MUSIC LM?
A team of engineers at Google is demonstrating a new generation music AI system called MusicLM. In an article published on the arXiv preprint server, the group claims the new system will establish new levels of composition and high fidelity for computer-produced songs.
2.HOW ITS TRAINED?
The system was taught to create music by training on songs played by humans for 28,000 hours. You can also create songs of variable length. For example, you can create a fast riff or an entire song. In addition, you can create moving pieces, often found in symphonies, to create a narrative atmosphere. The system can, for example, B. Request for a specific instrument or a specific genre.
3.WHY THE CREATION OF LM IS UNIQUE?
The development of MusicLM is part of a series of deep learning AI applications designed to emulate human mental abilities such as writing papers, drawing pictures, taking tests, speaking, and creating mathematical proofs. Several other efforts have been made to create song generation applications such as Dance Diffusion, Jukebox and Riffus.
Vocals and more accurate vocal noises can be generated on request, but the result tends to sound like a robot choir that doesn’t recognize the lyrics. Google does not release apps for general use. Tests show that about 1% of system-generated music is copied directly from human artists. Therefore, they are wary of content misappropriation embezzlement and lawsuits.