WanSong creates music like a sculptor carving a statue: it starts with noise, akin to the roar of a waterfall, and gradually chisels away the excess until pure notes and voices emerge. This method, diffusion, turns chaos into order, like ink dispersing in water, only in reverse. The model doesn’t stitch sounds together from fragments; it births a complete song in one go.
Unlike previous solutions, WanSong generates up to five minutes of music in a single pass and outputs two separate tracks: vocals and instruments. It’s a true foundational model that simplifies audio spectrum analysis for editing. On top of that, a special technique reduces the number of steps, making everything much faster.
The same diffusion principle was first taught to create photorealistic images. Now it composes full-fledged tracks, an unexpected pivot from visual to auditory creativity.
Diffusion neural networks first painted pictures, and now they compose music.
Just like in the movie 'Her,' where the OS composes music, WanSong brings that to life.