The paper introduces Stochastic Control Guidance (SCG), a novel method for symbolic music generation that leverages non-differentiable rule guidance. SCG is designed to work with pre-trained diffusion models, enabling plug-and-play guidance for rules that are often non-differentiable, such as note density and chord progression. The method is inspired by stochastic control theory and uses path integral control to find the optimal control policy that steers the diffusion process towards the target distribution. The paper also presents a latent diffusion architecture for symbolic music generation with high time resolution, capable of generating dynamic piano performances at 10ms intervals. Experiments demonstrate that SCG outperforms existing methods in terms of music quality and rule-based controllability, making it a powerful tool for composers and musicians. The framework is flexible and can be extended to other fields where non-differentiable constraints need to be enforced.The paper introduces Stochastic Control Guidance (SCG), a novel method for symbolic music generation that leverages non-differentiable rule guidance. SCG is designed to work with pre-trained diffusion models, enabling plug-and-play guidance for rules that are often non-differentiable, such as note density and chord progression. The method is inspired by stochastic control theory and uses path integral control to find the optimal control policy that steers the diffusion process towards the target distribution. The paper also presents a latent diffusion architecture for symbolic music generation with high time resolution, capable of generating dynamic piano performances at 10ms intervals. Experiments demonstrate that SCG outperforms existing methods in terms of music quality and rule-based controllability, making it a powerful tool for composers and musicians. The framework is flexible and can be extended to other fields where non-differentiable constraints need to be enforced.