In this work we present a novel approach for interactive music generation based on the dynamics of biological neural networks. We develop SANTIAGO, a real-time environment built in Pd-Gem, which allows to assemble networks of realistic neuron models and map the ac- tivity of individual neurons to sound events (notes) and to modulations of the sound event parameters (duration, pitch, intensity, spectral content). The rich behavior exhibited by this type of networks gives rise to complex rhythmic patterns, melodies and textures that are neither too random nor too uniform, and that can be modified by the user in an interactive way.