Please use a browser that is compatible with webaudio
loading...
Caching ~700mb model data. This will be (very) slow the first time, and then much faster next time you load this page.
This app will work best with a webgl2 compatible GPU.
CONCEPTULAR Beat Synthesiser
steps: step length (ms):
Model 1: Hihats
Model 2: Clave/Triangle
Model 3: Snare/Bongo
Model 4: Kicks
This is my message

Conceptular Beat Synthesiser is a drum machine powered by machine learning. There are no samples, instead the system uses neural network models of sounds.

To use, create a sequence using the grid, set the envelope parameters, and press the Render button. The rendering process needs to use the computation power of your GPU, and may take some time.

The real power of this synthesiser lies in the way you can manipule the sound models using the envelopes. Each model has the following parameters:

Morph: Each model is trained to reproduce two different sounds. You can morph between these.

Stretch: This envelope will change the sound to play more quickly or slowly, but will not change the pitch.

Leak Rate: This parameter changes the internal timing of the model. It tends to change the pitch of the sound but may have more unpredicable effects.

Excite: Raising this parameter will make each model behave in more unpredictable ways, while lowering it will make the sound softer.

Gain: This controls the volume of the sound.

The Share button will create a link to your sound that you can access another time, and also share with others.

The Download button saves the sound onto your computer.