Audio-reactive visualizer that actually reacts to the STEMS, not just the master's amplitude
Most 'audio-reactive' AI visualizers react to overall loudness, which means everything pulses together on the beat and it's boring by bar eight. The upgrade: split the track into stems (drums, bass, vocal, synth) and map each to a DIFFERENT visual parameter.
In mine: the kick drives a camera shake, the bass drives the scene's scale/zoom, the vocal drives a color temperature shift (warmer on the vocal, cooler without it), and the hi-hats jitter a particle layer. Now the visual has the same ARRANGEMENT as the song. When the vocal drops out, the color goes cold. When the bass hits, the world inflates. It reads as musical instead of just reactive.
Separating stems is trivial now with any stem splitter. The insight is to treat your visual parameters like a mixing board and assign each musical element its own channel. Routing diagram in a reply.