Conditional Image Translation is an emerging field within computer vision, continually providing new opportunities for innovation. This paper introduces a refined adaptation of the Pix2Pix model, emphasizing an enriched loss function that not only incorporates perceptual aspects for better semantic and structural representation but also leverages a meticulously researched, optimal weight ratio for the different components of the loss function. The specification of this weight ratio is determined based on the unique objectives of our research, marking a nuanced and tailored approach towards model optimization. To objectively measure the performance of our advanced model, we utilize the Fréchet Inception Distance (FID) as a reliable evaluation metric, ensuring precision and transparency in our assessment process. Upon testing on the facade dataset, our model exhibits a significant enhancement in generating high-quality, detailed images, thus verifying the effectiveness of our method in the broader domain of conditional image translation. The deliberate allocation of the weight ratio within the loss function represents a pivotal innovation, substantiating the importance of this research in the continuing evolution of the field.