AI tools

Humus treats AI as material to play with, not a chat window on the side.

Choose a model

Settings ▸ AI picks the provider: a model server on your own machine, the default, or a hosted service with your key. When the local server answers, a dropdown lists the models it has.

Presets from words

On any organism's preset strip, Evolve nudges the settings a little further each press. The strip's menu has Generate with AI: describe a sound and a preset is written for it, inside each knob's range. The Presets window adds Random and Evolve buttons. It all works on hosted plugins too.

The assistant

Control ▸ AI Assistant opens a conversation that can edit the patch: add and connect organisms, set knobs, change the tempo, drive the transport. Undo takes its changes back, and the conversation is still there when you reopen the window.

Sample Lab

Control ▸ AI Sample Lab designs a handful of sounds from a sentence, renders them to ~/Documents/Humus/Library/Samples/, and lays them across a Sampler so the keys play them. A card in the corner shows the progress.

SoundSpace

SoundSpace analyses a folder of sounds into grains and maps them by timbre. Fly a cursor across the map and a grain cloud plays whatever is under it. The cursor can be automated and mapped, so a path through the sound can be performed or sequenced.

Hands and Skeleton

Hands tracks your hands through the camera: each finger, the wrist, a pinch, and poses you teach it. Skeleton does the same for the whole body. Cord their MIDI outlet into an instrument, or right-click any knob and pick them under Control with.