With digital humans being used extensively in blockbuster movies, triple-A games, and the entertainment industry, the demand for digital humans is growing exponentially. In response, VFX studios are continuously expanding their research into 3D facial appearance digitization pipelines to generate realistic digital human facial appearances. However, existing 3D facial appearance digitization pipelines have limitations in terms of labor-intensive methods in R&D and system improvement through qualitative methods. Therefore, in this study, we propose a pipeline optimization of 3D face digitizing system through digital twins. The proposed pipeline predicts and solves problems in digitizing through extensive simulations in a virtual space that reflects the characteristics of the physical space with a digital twin and applies it to the physical space. It can solve the labor-intensive method, which is a limitation of the existing pipeline, through the interaction of physical and virtual space. In addition, by improving the system through quantitative evaluation of the facial appearance generated in the virtual space, the system can be optimized to generate better digital human facial appearance compared to the existing pipeline. The proposed digital twin pipeline can generate realistic facial appearance based on quantitative evaluation, and can be expected to reduce the cost of system construction. We believe that this research will contribute to the introduction and establishment of digital twin pipelines in the VFX industry.