Photorealistic style translation has gained significant attention in the field of computer vision and graphics due to its potential applications in many areas such as content generation, artistic expression and image editing. In this paper, we propose an improved hybrid Generative Adversarial Network (GAN) framework to perform photorealistic style translation. This model aims to overcome limitations observed in the existing literature by leveraging the strength of both GAN and Autoencoders. A fast and efficient feature extractor method is developed based on EfficientNet B2 to improve the performance of style transfer method. Moreover, two loss functions are proposed which further optimize the efficiency of the model in terms of accuracy and realism. We conduct extensive experiments on state-of-the-art datasets and demonstrate the effectiveness of the proposed model. Additionally, we provide analysis of results and discuss potential avenues for future research in the field of photorealistic style translation.