GenAI x Chinese Paintings

GenAI x Chinese Paintings is a research-creation project exploring the potential applications of artificial intelligence (AI) tools in the conservation and reconstruction of Chinese paintings. The project tests its methods using two case studies.

Case Study 1 focuses on a painting by the Qing dynasty artist Shitao, which contains a section suspected to have been inpainted by later hands. Using Stable Diffusion combined with a custom-trained Shitao-style LoRA (Low-Rank Adaptation), we virtually removed this section and generated multiple restoration proposals. The results demonstrate that the AI-generated inpainting seamlessly integrates with the surrounding composition and offers culturally and historically contextual options under guided prompts. This suggests that GenAI tools can provide valuable references for actual conservation work, addressing challenges such as a lack of skilled personnel, the risks associated with physical interventions, and conflicts with conservation restoration theories.

Case Study 2 builds upon a letter written by Shitao to the Qing dynasty artist Bada Shanren (currently housed in the Princeton University Art Museum). The letter specifies detailed requirements for a painting that has since been lost. Using the letter’s descriptions as a prompt, Stable Diffusion, combined with a custom-trained Bada Shanren-style LoRA, was employed to virtually reconstruct the painting. Additional limited refinements were made using Photoshop. The result faithfully adheres to the stylistic conventions of Bada Shanren while fulfilling the letter’s specifications, showcasing traditional Chinese landscape composition and the intellectual spirit of literati painting. The reconstruction process drew inspiration from the personal connection between Shitao and Bada Shanren, aiming to visualize their emotional bond. By including depictions of both artists in the reconstructed work, the project symbolically resolves their unfulfilled meeting in real life. This case further illustrates that AI tools can support not only virtual restoration but also speculative reconstruction and artistic creation.