The process, however, involves far more than entering a few prompts. The technology lowered the technical threshold, but it did not make the choices for him.

　　When creating Nantianmen, the mythical heavenly gate, Chen first defines the spatial rhythm and tonal hierarchy, pairing towering architecture and cloud seas with deliberately miniature figures to amplify the celestial realm's grand scale.

　　If the result does not match the picture in his mind, he tries again. He often generates over 1,000 images a day, revising a single scene repeatedly before moving on to animation, editing, color grading and music.

　　The machine can generate a thousand possible heavens. Chen still has to decide which one is his.

　　"AI-produced images can be soulless, and it takes human inspiration and aesthetic sensibility to endow works with warmth and vitality," he said.

　　Zhang believes Chen's experience reflects a broader shift brought about by generative AI. He said AI is giving people without conventional artistic training access to forms of expression that once required specialized skills. But what they create still depends on what they bring to the tool.

　　"A person's existing knowledge inevitably shapes how they use AI," Zhang said.

　　Chen brought his memories to the machine. Viewers abroad, in turn, brought theirs to the image.

　　"Everyone brings their own associations to these scenes," he said.

　　If his imagined world had to be reduced to three elements, he would keep clouds, architecture and people. The first two should feel immense, he said, while the last should remain small.

　　The details are culturally specific, but the visual effect is more immediate: tiny human figures dwarfed by vast clouds and monumental architecture. The responses suggest that viewers did not need to understand every cultural reference to find a way into the image.