无字书

Dialogues are generated in real time by a Large Language Model (LLM) during gameplay, instead of using pre-written fixed texts. Players need to input their own AI service API Key (DeepSeek is recommended) within the game. Dialogue content is generated and returned instantly by the selected service. The backbone of the story — chapter structure, key plot beats, multi-ending branches and ending judgment logic — is manually designed and built into the game by developers. The AI only handles real-time character wording, emotional responses, and the interpretation of player input. It does NOT determine the direction of the plot. The game has built-in content filtering and response quality validation. It will block non-compliant content and request re-generation. Players may switch AI models, adjust generation parameters, or turn off dialogue suggestions in settings. Speech synthesis and voice input are optional features; disabling them will not affect core gameplay. In addition, the primary art and audio assets (character portraits and animations, scene illustrations, background music) were created with generative AI tools during development, followed by manual screening, revision and post-processing. The game icon was also AI-generated and then manually refined.