A Diffusion Model-Based Character Synthesis Using a Blank Face

In this study, we propose a framework for synthesizing new characters by applying the features of a reference face to a source face using a diffusion model-based image editing technique. For an effective synthesis, a blank face of the sample is first generated by removing all features except the hairstyle, face shape, and skin tone. Then, facial features such as the eyes, eyebrows, nose, and mouth from the reference face are applied to the blank face of the source face. This strategy allows the creation of a new character that reflects the reference face’s features on the source face without producing unwanted artifacts or undesired blending of facial features.