Abstract

With the development of immersive technologies such as virtual reality, virtual experiences are becoming increasingly popular in fields such as education, healthcare and exhibitions. This places higher demands on the emotional adaptability and interactive expressiveness of content generation. The development of generative artificial intelligence (AIGC) technology has endowed virtual experience systems with the capability to generate efficient, multimodal content. This paper focuses on the 'emotion cue-driven AIGC generation mechanism', using 'anxiety–calm' as a representative emotion type. It then constructs an experimental framework covering both text and image tasks. The paper compares and analyses the generation performance of seven mainstream AIGC platforms. The study involves designing structured prompt templates and proposing a three-dimensional expert evaluation system comprising 'generative richness', 'emotional fit', and 'detail expressiveness', taking into account the contributions of language construction and visual rendering to virtual immersion. Experimental results show that Google Gemini and ChatGPT 4o demonstrate strong emotional expression and scene construction capabilities in text generation, while Midjourney excels in image generation tasks. It accurately responds to prompts and generates concrete, stylistically distinct visual content, making it suitable for emotion-driven virtual experience design. However, some platforms still exhibit significant gaps in detail representation and emotional consistency. This study proposes a fixed, prompt-driven mechanism based on emotion recognition which demonstrates cross-platform adaptability and virtual experience transfer value. It provides theoretical support and methodological pathways for the future application of AIGC in immersive education, virtual exhibitions and affective computing scenarios.

Download

Share

COinS