Current methods for handwritten Chinese text generation primarily focus on generating individual characters. However, for practical applications, users often require a handwritten text-line generation model to streamline their workflows. Generating stylized handwritten text-line necessitates capturing both character-level styles (e.g., stroke thickness, ink color, and cursive join) and text-line-level styles (e.g., inter-character spacing and text-line inclination). To address this challenge, we propose a novel cascaded diffusion architecture (i.e., Layout-Diff and Imitating-Diff) to disentangle these two levels of styles via customized template guidance. Specifically, Layout-Diff assists in synthesizing a template image, where the visual archetypes, rendered in a printed font, are well-arranged on a blank canvas. Subsequently, Imitating-Diff transforms this synthesized template into a stylized handwritten text-line image. We propose a latent aggregation module to efficiently incorporate textual guidance into Imitating-Diff while reducing computational cost. Furthermore, we design a high-frequency weighted loss for fine-tuning Imitating-Diff in the pixel space to preserve structure details with clarity. Extensive experiments demonstrate that our method effectively mimics diverse handwriting styles while ensuring structural accuracy in generated text lines.