Prompt Engineering and Semantic Mapping Details for Text-to-3D Scene Generation Using Tencent Hunyuan 3D and Tripo AI: A Case Study on the Science Fiction Literary Work Not in This Lifetime

This resource includes two supplementary technical appendices, centering on the research task of large language model empowered text-to-3D scene generation, with an emphasis on VR scene reconstruction for magical realism literary texts. The two documents are tailored for two mainstream text-to-3D frameworks, namely Tripo AI and Tencent Hunyuan 3D. They provide systematic pipelines covering prompt engineering design, semantic feature extraction, parameter mapping and reproducible generation workflows. A standardized end-to-end paradigm spanning text description, semantic parsing and 3D scene construction is established for academic research, which fully complies with the criteria of rigor, integrity and reproducibility required for technical appendices of high-level EI journals. Both appendices adopt ten curated chapters selected from the magical realism science fiction work Not in This Lifetime as the exclusive data source. Oriented to the core narrative scenarios of each literary chapter, this work constructs a complete technical loop consisting of prompt engineering formulation, intelligent semantic parsing, adaptive parameter transformation and model-based generation output. It ensures that the generated 3D scenes are highly consistent with the original contextual settings, emotional tones and inherent stylistic characteristics of magical realism. With standardized architecture and well-divided functional modules, the documents can be directly applied to research and engineering practices including literary text-driven VR/3D content creation, digital heritage scene reconstruction and immersive narrative visualization. The first appendix entitled Prompt Engineering for Text-to-3D VR Scene Generation with Tripo AI (A Case Study of Not in This Lifetime) is specially customized for the Tripo AI platform. It adopts a five-module structured prompt template that covers scene contextual description, core semantic extraction, aesthetic constraint definition, geometric boundary specification and standardized output formatting. Supported by excerpts from the original literary text, complete prompt samples and GPT-4V analytical outputs are provided to define quantifiable metrics, including primary and secondary semantic entities, RGB color proportions, surface roughness, illumination color temperature, spatial dimensions and topological relationships. Corresponding parameter conversion tables are also formulated to standardize the transformation from natural language narratives and LLM analytical results to Tripo AI-compatible executable parameters. The covered items include model selection, style tagging, geometric specification configuration, rendering modes, vertex quantity constraints and export formats, enabling direct API invocation and high-fidelity one-click 3D scene generation. The second appendix named Prompt Engineering and Semantic Mapping Details for Text-to-3D Scene Generation Based on Tencent Hunyuan 3D is adapted to the Tencent Hunyuan 3D model. While maintaining the original prompt framework, inherent scene semantics and magical realism expressive logic, targeted adaptive optimization is implemented for model configuration, texture synthesis, light simulation and spatial layout constraints. The standardized outputs contain scene overviews, core constituent elements, aesthetic attributes, geometric restrictions, magical realism expression modes, generation priority schemes and parametric suggestions. This forms a cross-model standardized workflow, allowing researchers to rapidly migrate technical schemes and conduct comparative experiments across different 3D generation platforms. Collectively, the two appendices establish an engineering-oriented reproducible pipeline for converting literary narratives into 3D scenes. Firstly, a structured prompt specification oriented to literary narration is developed to precisely extract architectural landscapes, natural environments, functional facilities, emotional atmospheres and magical realism exclusive elements, and to clearly distinguish tangible modeling entities from abstract artistic conceptions. Secondly, quantitative mapping from textual semantics to 3D engine parameters is realized, offering reusable standard configurations for color, texture, illumination, scale, orientation and structural layout. Thirdly, the scheme balances general adaptability and model-specific characteristics; the template framework can be extended to analogous literary works by fine-tuning the customized magical realism module. Fourthly, the solution strictly adheres to original narrative settings to completely preserve and faithfully restore key features such as spatial layout, architectural morphology, textural vicissitude, nostalgic ambience and the immersive perception of temporal stagnation. This resource can serve as experimental appendices, technical guidelines and reproducible benchmarks for digital humanities, computer graphics, immersive media and VR content generation research. It enables researchers to efficiently reconstruct core 3D scenes corresponding to ten independent chapters of Not in This Lifetime. Furthermore, the embedded prompt templates and semantic mapping logic can be directly reused to conduct broader research on text-driven 3D scene generation. The resource satisfies academic publishing requirements for open data and reproducible workflows, adheres to the FAIR data principles of the Zenodo platform, and facilitates citation, secondary reuse and result validation among academic peers.