Importing Diffusion and Neural Styles for Realistic Face Generation
Tang-Wei Su, Jhe-Wei Lin, Cheng-Hsuan Lee, Wei-Ming Tseng · 2023
Face generation has made significant progress with large-scale diffusion models, attracting widespread interest. The success of Generative Adversarial Networks (GANs) in face generation is particularly notable. We developed a new approach of diffusion modeling combined with neural style transfer for realistic face image generation. We applied the concept of "style loss" from neural style transfer to maintain the original style features while transforming random photos into face images. This integrated approach provides more artistic and personalized results while maintaining image fidelity. The result has far-reaching impacts and improves face synthesis techniques in art and entertainment. Further research is needed to realize this potential.