Neural audio codec (autoencoder) is a module used in many AI music generation systems. By unpacking a codec, one can directly interact with the sound generation process, and build tools that link to sensors, hardware, gestural controllers... We introduce Latent Terrain, a Max/MSP tool package designed to tailor codecs into corpus-based sound spaces, to build AI musical instruments.

We present an annotated portfolio resulting from a collaborative artistic exploration with four NIME makers, to showcase the design space opened by autoencoders. Reflecting on our practice-based account, our paper discusses the challenges and opportunities of enabling neural audio autoencoders as design materials for AI-enhanced NIMEs.