Abstract: Digital humanities research has traditionally centered on textual analysis, yet this "text-centrism" paradigm reveals significant limitations within the Chinese context, including insufficient character set coverage, low OCR accuracy, and the loss of non-textual cultural information, all of which hinder a comprehensive study of China's rich material cultural heritage. The emergence of multimodal technologies offers a transformative pathway for Chinese digital humanities. This paper investigates the predicaments of text-centrism, analyzes solutions enabled by multimodal fusion technologies, and uses DeepSeek’s Janus Pro model as a case study to illustrate the potential of unified multimodal large-scale models in ancient text digitization, intelligent agent development, and cultural heritage preservation. The results show that multimodal technology can reconstruct cultural memory through cross-modal synergy, enhance the public's cultural identity, and provide technical and methodological support for the transformation of Chinese digital humanities.

Key words: Digital humanities, Text-centrism, Large multimodal models, DeepSeek, Multimodal technology