Capturing high-quality images under low-light conditions remains a significant challenge, particularly for mobile cameras, which have relatively small sensor sizes compared to DSLR cameras. A promising approach to enhance low-light imaging involves fusing information from two consecutive captures with different exposure settings: a long-exposure image, which is less noisy but often blurred, and a short-exposure image, which is sharp but noisy. Recent advances in deep learning (DL) have demonstrated significant improvements over classical methods. However, these approaches require large-scale datasets for training. Collecting real-world data is technically demanding and laborious, leading prior work to rely on synthetic data. Existing methods either generate blur using video frames, thus limiting the diversity of blur to the camera motion, or simulate blur-noise pairs from still images, assuming planar scenes and neglecting depth variations. In this paper, we propose a pipeline that leverages 3D Gaussian Splatting (3DGS) to simulate camera motion within the 3D scene, allowing for the rendering of realistic blurry and noisy image pairs. We compare the synthesized data to real-world images and demonstrate the fidelity of our approach. This work lays the foundation for generating large-scale datasets suitable for training DL models aimed at low-light image enhancement.