nsendu-flash by nnirror

nsendu-flash was created by live-coding the latent space of a neural audio synthesis model (via RAVE and nn~) trained on my previous albums.

I created most of the material for nsendu-flash with a Max patcher that allows for deterministic movement through permutations of the model's 8 latent space parameters. This iterative movement creates rhythmic patterns as it moves across all permutations. I added controls to configure a subrange of the latent space to move through, a speed for movement, and a degree of quantization for the latent parameters.

The patcher leverages how variational autoencoders produce output variations given the same latent space input, creating an interplay between the deterministic latent space movement and the probabilistic autoencoder behavior. I improvised between these two poles by live coding parameters with Facet.

The remaining material for nsendu-flash comes from other novel configurations of nn~ objects within Max, using the same model.

Equipment used: IRCAM RAVE, nn~, Max, and Facet.

Production and mastering by nnirror.

designing and exploring generative music systems.

nnirror.xyz
post.lurk.org