Photo Credit: Mike Castro Demaria
Google researchers have created an AI capable of generating music from text prompts—like ChatGPT. The model is called MusicLM.
A text prompt like ‘a calming violin melody backed by a distorted guitar riff’ will generate music at 24kHz for several minutes. The paper published by researchers says MusicLM outperforms previous music generation models both in audio quality and adherence to the text prompt. Researchers say MusicLM can also be trained on both text and a melody—meaning it can transform whistled and hummed tunes to the style from the text prompt.
Google has also published some music samples that MusicLM has generated based on text prompts, just like ChatGPT. Some of the prompts Google has used to generate music include:
- The main soundtrack of an arcade game. it is fast-paced and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds, like cymbal crashes or drum rolls.
- A fusion of reggaeton and electronic dance music, with a spacey, otherworldly sound. Induces the experience of being lost in space, and the music would be designed to evoke a sense of wonder and awe, while being danceable.
What’s notable about MusicLM is that not only can it seamlessly combine genres and instruments—it grasps abstract concepts like ‘catchy’ and ‘otherworldly.’ MusicLM also features a story mode, that is capable of stitching several descriptions together to create a soundtrack that evokes certain feelings. There are limitations to the current model. It doesn’t handle vocals well at all and it tends to rely on repetition more than human-created music would.
Google researchers have been experimenting with artificial intelligence models for years. MusicLM won’t be released to the public due to copyright concerns. There’s currently no standard for licensing AI music trained from copyrighted works. MusicLM isn’t being released to the public, but it does provide an example of what’s capable for music generation.