Audio Lifeforms

It's like the gantz graf video, how are they creating this?

it's fairly easy from a technical POV, modern 3D animation software has basically all the parts ready to click together with a mouse.

create or load some 3D model to taste, then there's a variety of geometry modifier effects which can be automated over time. slap a bunch of them on your 3D model. then you hook up some of the effect's parameters to audio loudness. the audio is likely split into a bunch of frequency bands with FFT so it reacts differently to bass than high frequencies. the rest is a bit of tweaking.

tbh this looks a bit nice but it also looks a bit lazy, it's still quite apparent that a lot of detailed manual work went into Gantz Graf, not so much here. looks like the person who made this used a little bit of all the effects / shaders he had at his disposal.

I'm doing stuff with this in Puredata's GEM renderer. A digitized audio signal is given a numerical value and you can scale the control data to go into vector coordinate(s) in a 3d object to control the x/y/z parameters to transform, rotate, scale and reflect geometry. The numerical value can be just the average amplitude of the track, individual amplutide signals of different tracks and sounds, or even mapping different functions to frequency segments.