Time Tunnel to the past: Generative AI as an approach for visualizing ancient Chinese auroral records

The extreme magnetic storm of September 1770 stands as a landmark event in space weather history, with auroral displays recorded across East Asia that rival the Carrington Event in magnitude. However, the visual reality of this phenomenon remains locked within the abstract, metaphorical descriptions of ancient chronicles. This study introduces a comprehensive omni-modal reconstruction framework that leverages Generative Artificial Intelligence (AI) to “revive” these lost celestial scenes. We propose a systematic pipeline integrating Text-to-Image, Image-to-Image, and Text-to-Video generation to transform the textual records of the Wenxian Tongkao into high-fidelity visual assets. Unlike purely artistic renditions, our methodology incorporates a physical constraint layer—derived from ionospheric physics—to guide the AI, ensuring that the reconstructed “red light spanning the sky” over Dongting Lake is both aesthetically evocative and scientifically plausible. This work demonstrates that Generative AI can serve as a “Time Tunnel,” bridging the gap between historical text and modern visualization, and offering a novel tool for digital heritage preservation and public science communication.