Lessons from 200 demo uploads

Everyone uploads short clips

Median was 38 seconds. Makes sense—people want a quick yes or no before they commit to anything.

Means the first 20 seconds really matter. That's where we focus when tuning how fast the vocal stem settles in.

What people actually test

Talking heads and singing hooks, mostly. Makes sense—those are the clips where a clean vocal matters most. Works best when the voice is front and center in the mix.

We also see a fair amount of location audio. Interviews shot outside, vlog footage, that kind of thing. People want to know if they can save a take before they commit.

The pattern

Strong vocal in = clean stem out. When the voice sits clearly in the mix, the residual stays musical and the isolated track sounds natural. That's what we're optimizing for.