High-fidelity Facial Exchange: Exploration and Application of Model Innovation and Super-resolution

Face swapping technology has specific and important application prospects in the age of artificial intelligence, and with the development and advancement of computer technology, more and more researchers are investing in this area of research. However, previous research results are difficult to synthesize quickly with high fidelity while focusing on the high quality of the face. This study proposes a high-fidelity and realism synthesis method based on traditional model innovation while combining facial super-resolution techniques. Specifically, we first optimise and improve the original algorithm, and after using the model synthesis, the model is enhanced frame by frame using a face facial image recovery model for facial resolution, after which the results are combined for video. At the same time, we propose a subjective evaluation method based on the quality of facial generation, further assessing the effect of the synthesis quality by combining objective and subjective aspects. Through experimental comparison, this implementation outperforms the more mainstream models in terms of quality and efficiency, providing new ideas and methods for the development of facial synthesis technology.