Google’s freakish AI music generator, MusicLM, is capable of creating entire songs and sounds from text.
The artificial intelligence music creation renaissance has arrived.
There’s nothing we’re able to do to stop it, but we can certainly embrace it. Google’s freakish new MusicLM program uses AI to generate high-fidelity music from text prompts.
The company has now publicly released MusicCaps, a dataset consisting of 5,500 music-text pairs and trained on 280,000 hours of music, to support future research. The MusicLM AI was tested by humans, who fed the program a series of text prompts to analyze the accuracy of its music generation.
In one example, researchers entered a rich text prompt that read, “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.”
When generating longer music in various styles, researchers kept the prompts short. Here’s what MusicLM generated when researchers prompted it to produce a “melodic techno” track.
Major tech companies are scrambling to embed AI into their infrastructures and develop programs to enhance the user experience. While this new tech is still quite nascent, it’s only going to get better, more accurate and fiercely competitive.
Last year, we asked ChatGPT questions about EDM, and it knew a lot more than we thought it would. Highly nuanced questions, like the difference between Excision and Rezz’s respective sounds, were answered with clarity and accuracy.
On the visual side, the EDM.com team used an AI image generator to create fanciful images of DJs.
Lennon Cihak
Source link
