This study specifically explores the use of Neural Radiance Fields (NeRF) methodology to generate volumetric representations of three-dimensional environments, starting from two-dimensional images derived from photographic captures and video frames. This approach reconstructs the geometric and visual properties of the acquired scene with high fidelity. The NeRF methodology is complemented by Gaussian Splatting, which leverages volumetric Gaussian distributions to describe the visual and perceptual properties of scene components, effectively preserving the specular and reflective behavior of materials. This research underscores the potential of these methodologies as a major step toward achieving more realistic and adaptive virtual reality applications, particularly in the reproduction of architectural spaces. Nonetheless, challenges remain, such as optimizing the pipeline for even more intricate environments and adapting the models for large-scale applications. Looking ahead, the integration of these technologies could revolutionize the design field, enabling architects and designers to interactively and immersively explore projects still in development. This interdisciplinary approach, combining deep learning, computational graphics, and architectural science, highlights the potential to redefine not only how we experience virtual reality but also how we design and interpret the built environment.