This paper introduces DiLightNet, a novel method for fine-grained lighting control in text-driven diffusion-based image generation. The method consists of three stages: (1) generating a provisional image under uncontrolled lighting, (2) resynthesizing the foreground object using radiance hints and a refined diffusion model (DiLightNet), and (3) inpainting the background to match the target lighting. Key to DiLightNet is the encoding of the provisional image and multiplying it with radiance hints, which helps retain the shape and texture information while guiding the diffusion process. The method is evaluated on various text prompts and lighting conditions, demonstrating its ability to produce plausible images that match both the text prompt and the target lighting. The paper also includes an extensive ablation study to validate the effectiveness of each component of DiLightNet.This paper introduces DiLightNet, a novel method for fine-grained lighting control in text-driven diffusion-based image generation. The method consists of three stages: (1) generating a provisional image under uncontrolled lighting, (2) resynthesizing the foreground object using radiance hints and a refined diffusion model (DiLightNet), and (3) inpainting the background to match the target lighting. Key to DiLightNet is the encoding of the provisional image and multiplying it with radiance hints, which helps retain the shape and texture information while guiding the diffusion process. The method is evaluated on various text prompts and lighting conditions, demonstrating its ability to produce plausible images that match both the text prompt and the target lighting. The paper also includes an extensive ablation study to validate the effectiveness of each component of DiLightNet.