Neural activity and behavior in awake animals reflect the interaction between multiple fast processes, such as movements, and slow processes, including arousal, neuromodulation, and learning. A variety of experimental and modeling approaches have successfully tackled single aspects along this spectrum, but a comprehensive multi-scale framework is lacking. To address this challenge, we developed two-photon (2p) mesoscope assays to record neural activity at single-cell resolution from the dorsal and lateral cortex of freely running mice during behavior. We found that whole-brain single-cell dynamics unfold through complex dynamical activity patterns, distributed across many areas and varying over multiple timescales simultaneously. To disentangle the latent dynamical factors underlying these dynamics, we deployed factorial hidden Markov models (fHMM) to capture the variability in neural activity as well as behavior. The fHMM revealed that whole brain dynamics arise from the superposition of latent dynamical factors, each factor encoding specific aspects of behavior and distributed across the cortical hierarchy according to a high-dimensional compositional logic. Latent dynamical factors were interpreted based on their switching kinetics as encoding a variety of behavioral variables over multiple timescales, including movements of independent body parts, slow variations in arousal level, and longer periods of activity/inactivity. Single excitatory neurons exhibited mixed selectivity to multiple latent factors across timescales from hundreds of milliseconds to tens of seconds, and the expression of latent factors showed specific spatial organization across dorso-lateral and sensory-motor axes of neocortex. Overall, our fHMM elucidates the computational principles underlying whole-brain dynamics during behavior, by disentangling the multi-scale and distributed compositionality of latent neural factors encoding behavioral variables and internal states, giving rise to the complex yet well-organized symphony of the dynamical brain.