以人工智慧繪畫輔助虛擬角色之美術設計

Art Design of Virtual Characters with AI Painting Assistance

葉冠麟(Kuan-Lin Yeh)；李俊賢(Jiunn-Shyan Lee)

人工智慧繪畫 ； Stable Diffusion ； 角色美術設計 ； AI Painting ； Stable Diffusion ； Character Art Design

修平學報

48期（2024 / 09 / 01）

23 - 42

繁體中文;英文

本研究主要在探討發想虛擬角色美術設計的過程中，導入人工智慧繪圖技術來輔助繪製之可能性。虛擬角色的美術設計是電子遊戲和動畫製作中的一個重要環節，它決定了角色的視覺概念與美術風格方向。然而傳統的角色美術設計需要花費大量的時間和精力去產出，並且受限於人類的想像力和創造力。因此筆者尋思藉由開源AI繪圖軟體：Stable Diffusion的輔助，建構一個有效率的角色概念設計圖之生產流程。本研究使用開放原始碼的AI繪圖軟體：Stable Diffusion為輔助工具，內容包括以文生圖（Text to Image）、ControlNet、以圖生圖（Image to image）、內補繪製（Inpaint）等相關技術。此方案可以幫助設計工作者因應各種主題的設計需求，快速產出虛擬角色之三視設計圖，且創造出大量不同造型的角色，可以提高工作效率，並且提供更多的創意靈感和可能性。

This study aimed to explore the possibility of introducing artificial intelligence painting technology to assist in the process of art design of virtual characters. The art design of virtual characters is an important part of electronic game and animation production, which determines the visual concept and artistic style direction of the characters. However, traditional character conceptual design requires a lot of time and effort to produce, and is limited by human imagination and creativity. Therefore, the author seeks to use the open-source AI painting software: Stable Diffusion as an aid to construct an efficient production process for character conceptual design drawings. This study uses open-source AI painting software: Stable Diffusion as an auxiliary tool, including related technologies such as Text to Image, ControlNet, Image to image, Inpainting, etc. This solution can help designers respond to various design needs and quickly produce three-view design drawings of virtual characters. It can create a large number of characters with different styles, improve work efficiency, and provide more creative inspiration and possibilities.