This week, I moderated a panel at the Local Media Consortium with Annika Toernqvist, Ian Hill, and Robert Young entitled “How Publishers are Using AI Now“. We opened with a musical number I created with Udio called “Let’s Talk About… AI!” Here’s the song:
The prompt used to create the song was “A 1950’s pop song in the style of the Ronnettes produced by Phil Spector, big band, jazzy, female vocalist, bright vocals Female vocalist, Pop, Jazz, Swing, Traditional pop, Big band”
Here’s the song on Udio: https://www.udio.com/songs/pjQfXqHy6PfYbNkZpVAo8W
Udio (to my understanding) uses a diffusion model for creating its sounds. It’s a fascinating and unintuitive process. Here’s a good overview (https://towardsdatascience.com/audio-diffusion-generative-musics-secret-sauce-f625d0aca800) but before reading, I recommend you watch this video about image diffusion first (below). Once you know the principle of denoising static into images, it’s a lot easier to understand denoising waveforms into songs! It’s a remarkable (and very human) feat of ingenuity:
Leave a Reply