Google DeepMind

Lyria 3 is Google DeepMind’s most advanced AI music generation tool to date. Developed in collaboration with producers and musicians, it has a deep understanding of musicality – from rhythm to arrangement.

The platform empowers users to express, explore, and experiment with high-fidelity music, using simple prompts to create tracks with a natural, note-to-note flow. Users simply describe their desired track and dial in the specific details, while Lyria 3 handles the underlying complexity, putting entirely new musical possibilities at their fingertips.

In close collaboration with the Google DeepMind team and Art&Graft, we created a dynamic system of audio-reactive music visualisers for use on Google DeepMind’s website, helping to explain and visualise the capabilities of Lyria 3.

Each visualiser reacts in real-time to a song's pitch, tempo, and other characteristics to create a custom graphic that matches the sound. Visually, it appears as a soft, glowing gradient that pulses and shifts with the music. As it moves, the four signature Google colors shine through the glow, creating subtle, iridescent refractions.

We expanded the system's functionality to process image-based inputs. The resulting graphics dynamically illustrate the real-time transformation of static imagery into audio waves.