This paper introduces a novel AI-aided architectural design method driven by daylight, enabling end-to-end generation from massing models to architectural designs. The method combines massing model generation, daylight-driven façade design, and text-to-image generation using diffusion models. The process begins with generating massing models through parametric algorithms, which are then refined using a daylight-driven strategy to determine optimal window layouts. These models are then used to generate architectural designs by integrating large-scale language models with text-to-image models, enhancing the efficiency of visual design rendering. The massing model generation process involves creating initial cubic models and modifying them through addition and subtraction operations. These models are generated using the Rhino-grasshopper platform, allowing for precise control through a visual interface. The daylight-driven strategy involves creating a dataset of daylighting maps using solar irradiation parameters and Grasshopper software. These maps are then used to train a LoRA model for generating accurate daylighting maps. The trained model is then used to generate architectural designs by integrating with diffusion models and ControlNet for precise architectural design control. The architectural design generation process involves using text prompts generated by large-scale language models such as GPT-4 to create architectural design proposals. These prompts are then fed into diffusion models to generate visual architectural designs. The method has been tested and shown to effectively assist architects in quickly conceptualizing and creating initial design proposals, providing a new pathway for architectural design development. The method's main contributions include a novel AI-aided architectural design method driven by daylight, the first incorporation of daylight factors into deep learning-based architectural generation methods, and the seamless integration of large-scale language models with text-to-image models to enhance the efficiency of generating visual architectural design renderings. The method has been validated through experimental results, demonstrating its effectiveness in improving design efficiency and providing practical support for architects' creative endeavors.This paper introduces a novel AI-aided architectural design method driven by daylight, enabling end-to-end generation from massing models to architectural designs. The method combines massing model generation, daylight-driven façade design, and text-to-image generation using diffusion models. The process begins with generating massing models through parametric algorithms, which are then refined using a daylight-driven strategy to determine optimal window layouts. These models are then used to generate architectural designs by integrating large-scale language models with text-to-image models, enhancing the efficiency of visual design rendering. The massing model generation process involves creating initial cubic models and modifying them through addition and subtraction operations. These models are generated using the Rhino-grasshopper platform, allowing for precise control through a visual interface. The daylight-driven strategy involves creating a dataset of daylighting maps using solar irradiation parameters and Grasshopper software. These maps are then used to train a LoRA model for generating accurate daylighting maps. The trained model is then used to generate architectural designs by integrating with diffusion models and ControlNet for precise architectural design control. The architectural design generation process involves using text prompts generated by large-scale language models such as GPT-4 to create architectural design proposals. These prompts are then fed into diffusion models to generate visual architectural designs. The method has been tested and shown to effectively assist architects in quickly conceptualizing and creating initial design proposals, providing a new pathway for architectural design development. The method's main contributions include a novel AI-aided architectural design method driven by daylight, the first incorporation of daylight factors into deep learning-based architectural generation methods, and the seamless integration of large-scale language models with text-to-image models to enhance the efficiency of generating visual architectural design renderings. The method has been validated through experimental results, demonstrating its effectiveness in improving design efficiency and providing practical support for architects' creative endeavors.