How to Combine AI & Manual Work to Create Beautiful Digital Portrait

"I have been exposed to CG art for three years. In school, I studied traditional sculpture, and there were no specialized CG courses. I accidentally saw some content about CG art on the internet and was very interested. After that, I improved my professional abilities by purchasing online and offline courses, YouTube videos, and doing a lot of exercises."

"My goal in making this character is to independently complete a full-process digital human character and optimize my modeling workflow, design process, and asset availability. And making a character like this would round out my portfolio even more.

First of all, before starting the project, I collect a lot of reference materials to provide me with inspiration and ideas, especially for this type of original character."

"During the design process, I used a lot of Stable Diffusion to make my drafts more polished. This tool is really great for providing inspiration to creators. I just made a rough sketch in Photoshop and put it in SD. Let it compute so it can produce a lot of pictures in a short time. I choose the design that fits my own ideas among these pictures and continue to return to Photoshop for optimization. This operation is repeated many times until I get a satisfactory design.

For the head and body, I completed the modeling in ZBrush. I started with a sphere using DynaMesh. In terms of creation, I like to use a sphere as the initial basic model, which allows me to create more freely. During this process, I didn't need to think about the model's mesh. I didn't have to be too careful when making it. I can make it very easy and fast drawing a rough image of the model. Once the rough image was settled on, I used Wrap4D to give the model a MetaHuman mesh. For the production of high-end models, I used some scanned materials, such as materials provided by TexturingXYZ, 3D Scan Store, Triplegangers, and other platforms, and then matched and optimized them according to the characteristics of the character."

"Since I was making an image of a young Asian woman, her structure was very subtle and the model went through many iterations, but the end result was satisfying.

To create the clothing, I used Style3D, which is a very powerful clothing-creation software. I was blown away by its retopology capabilities. When making clothing in the past, manual topology was often involved. However, the topology function of this software is very powerful, and it can provide me with a grid that perfectly meets the project requirements without much manual operation.

"I also use a lot of artificial intelligence to give myself feedback during the production process. The AI ​​calculations for hair are very satisfying. I roughly determined the shape of the hair and soon threw in Steady Diffusion, which can give me a lot of amazing hair detail. After repeated iterations, a satisfactory hairstyle was finally achieved."

"The renderers I used for this project were V-Ray and Unreal Engine 5. This is my first encounter with V-Ray. I really like its SSS effect, and there are a lot of adjustable parameters so I can control each effect very accurately. This is also my first time using UE5. I'm not familiar with it yet. There is a lot for me to continue learning. I hope the effects of UE5 can be improved in the future. I look forward to it."

"This character was created in my spare time and took 4 months. This is also the result of my learning and experience during this period. In this process, I gained a lot of new knowledge and learned many techniques. In the days to come, I will continue to create better quality works for myself and everyone else."