Google researchers have created an AI capable of generating music from text prompts, such as ChatGPT. The model is called MusicLM.
A text prompt like “soothing violin melody accompanied by a distorted guitar riff” will generate music at 24 kHz for several minutes. A paper published by the researchers states that MusicLM surpasses previous music generation models both in sound quality and in following a text prompt. The researchers say that MusicLM can also be taught both text and melody — meaning it can transform whistling and humming melodies into a text hint style.
Google has also published some music samples created by MusicLM based on text prompts, as has ChatGPT. Here are some of the tips that Google used to create music:
The main soundtrack of the arcade game. he is dynamic and optimistic, with a memorable electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds such as cymbal beats or drum roll.
A fusion of reggaeton and electronic dance music with a cosmic, otherworldly sound. It evokes a sense of loss in space, and the music is designed to evoke a sense of wonder and awe, while being danceable.
What is remarkable about MusicLM is that it not only can easily combine genres and instruments, but also covers abstract concepts such as “memorable” and “otherworldly”. certain feelings. The current model has limitations. He’s generally bad at vocals and tends to rely on repetition more than music created by humans.
Google researchers have been experimenting with artificial intelligence models for many years. MusicLM will not be released to the general public due to copyright issues. Currently, there is no standard for licensing AI music created on the basis of copyrighted works. MusicLM is not released to the general public, but it is an example of what music generation is capable of.
