AI Music and Sound Production

This hands-on demonstration looks at how AI is entering the everyday workflow of a sound designer, and where it still needs a human ear.

Part one uses Krotos Studio Pro, a sound design tool built on professionally recorded, licensed sound libraries rather than generated audio. We will explore a new kind of foley: instead of placing samples on a timeline, sounds are performed in real time with a mouse or MIDI controller, like playing an instrument. We will build ambiences from a text description or a single image, then use Krotos Video to Sound inside Adobe Premiere Pro, where AI analyses the footage and proposes a synchronised soundtrack scene by scene. Each result is then shaped, performed and mixed by hand: the AI proposes, the sound designer decides.

Part two moves from tools to method. Drawing on The Sound of Archetypes, my practice-based research project presented at CMMR 2025, I will show how an idea can be turned into sound through language. A written set of rules describing a symbolic sound world is given to Claude (Anthropic), which builds the session in Ableton Live through an open-source MCP connection, creating tracks, loading devices and proposing parameter ranges. Then comes the part AI cannot do: listening. Using a single control mapped to many parameters, I will compare the AI’s first proposal with my ear-tuned version, so the audience can hear the gap between knowing about perception and actually perceiving.

The session is aimed at sound designers, filmmakers, composers and educators who want practical workflows rather than hype, and an honest view of where AI helps, where it falls short, and why the ear remains the final judge. The research continues towards real-time, interactive and immersive installation work.