Magenta RealTime Show & Tell: Crafting New Tools to Craft New Sounds

Musical instruments shape the music we make. So how can we bend generative technologies to our will and build the instruments of our dreams? Come see some research previews of the many different directions we’re taking real-time music generation in: from generative loopers and audio-semantic neural bending to KV cache manipulation, latent space visualizations, and audio-reactive synthesis, we’ll try to blow open the boundaries of neural music synthesis. Throughout the evening, part show & tell, part conversation, we’ll first build some technical understanding of Magenta RealTime 2, a state-of-the-art open and local live music model, and then discuss the creative affordances unlocked by real-time audio generation. There will be demos, co-design sessions, and an open conversation on how we can imagine a different unfolding of AI in the music world, where technological progress means more diverse forms of expression, while humans (sloppy or otherwise) remain center stage. Optional, but you’re welcome to bring your laptop and any other sound-making devices. No prior coding experience…