Kilnwave
A generative audio engine that writes stems, not loops. Describe the room and it returns a session you can open, split and re-voice.
Every colour on this page is light the browser generated, not an image somebody shot.
The spectrum is the interface.
Drag any band, the rest of the take is untouchedNothing here is retrieved. Every stem is written the moment you ask for it, which is why it is yours.
Built for rooms that ship daily.
Measured on the shared cluster over the last 90 days, not on a reference machine.
Frequently asked
Is the output actually licensable?
Yes. Nothing is retrieved from a training corpus at generation time, so every stem you export is yours outright, with a signed provenance manifest attached to the session.
Can I bring my own stems in?
Upload up to eight references per session. They condition the take without being reproduced, and the context lock will hold their key and tempo across regenerations.
What happens to an open session offline?
Sessions are local-first. The field, the stem split and the last twelve regenerations stay on your machine, and only the brief is sent when you ask for a new take.
Does it run on a laptop?
The editor does. Generation runs on the cluster, and a session streams its first audio in about 38 ms, so the laptop only ever handles playback and the split.