Research
& Development
WHAT WE DO
Foundational Model Development
We investigate the architecture of sound. By training foundational machine learning models from the ground up, we map the intricate details of timbre, spatiality, and performance dynamics. This process yields flexible, organic structures that expand the boundaries of traditional audio synthesis.
Neural Audio Productization
We translate complex computational research into playable, tactile instruments. By softening the edges of neural networks, we build intuitive interfaces that allow experimental processing to fold naturally into traditional studio workflows.
API & Licensing of our technology
We collaborate to optimize inference workflows, embedding advanced computational models within diverse hardware and software ecosystems. From cloud-based processing to local, real-time execution, we structure precise, technical pathways for seamless audio implementation.
Custom Collaborative R&D
We work alongside artists and organizations to cultivate proprietary tools and bespoke model architectures. By navigating the structural complexities of neural audio processing, we function as an integrated research partner for custom development, training, and sound design.
PARTNERS
PAPERS
Neutone SDK: An Open Source Framework for Neural Audio Processing
→The Neutone SDK is an open-source Python framework that streamlines the integration of PyTorch neural audio models into DAWs by managing the structural complexities of real-time processing.
Accepted to AES International Conference on Artificial Intelligence and Machine Learning for Audio 2025
Latent Granular Resynthesis
→A novel technique for creative audio resynthesis that operates by reworking the concept of granular synthesis at the latent vector level of neural audio codecs.
Accepted at ISMIR 2025 Late Breaking Demos