As a novice 3D model artist, one of the first difficulties I encounter is creating a good-looking human model. This challenge arises due to several reasons: unfamiliarity with human anatomy, difficulty in identifying flaws without proper shading on the raw model, and a lack of high-quality reference pictures.
By utilizing the ControlNet function, I can improve my workflow significantly. For instance, I can feed a screenshot of the raw model into ControlNet using OpenPose or line art edge detection to create my own reference pictures. Initially, I used a VRoid raw model to generate these reference images. After I create my own model, I can also feed a screenshot of it into ControlNet to visualize how it might look with textures applied. This helps me determine if my model is accurate.
Another way I incorporate generative AI into my 3D modeling is for creating textures. Face and eyes, for example, always require UV mapping and painted textures. I can feed a screenshot of the raw face model into ControlNet to generate a detailed face picture. Then, I can map the UV coordinates of the eyes and face onto the generated image.
