Character premise角色设定
Negative emotions are personified as collectible toy-like characters with distinct roles and moods.把负面情绪人格化为可收藏的玩偶角色，每个角色有明确身份和情绪。
AIGC Visual Product DirectionAIGC 视觉产品方向
BugBox
An emotion-led character IP system exploring how AIGC can shape a repeatable visual product direction. 一个以情绪角色为核心的 IP 视觉系统，探索 AIGC 如何形成可延展的视觉产品方向。
01 — The Brief01 — 项目命题
A character system that could grow beyond one generated image.不是一张生成图，而是一套可延展的角色系统。
BugBox is an AIGC concept project for an emotion-led character IP system, exploring how role settings, style anchors, and controlled generation can form an extendable visual product direction. BugBox 是一个围绕情绪角色与可生成 IP 视觉系统的 AIGC 概念项目，重点验证如何通过角色设定、风格锚点和生成控制，形成可延展的视觉产品方向。
Style consistency风格一致性
The case tests whether different characters can share material, lighting, proportion, and art-direction rules.项目测试不同角色是否能共享材质、光影、比例和美术方向规则。
Scene expansion场景扩展
Generated scenes extend each character into a small narrative world without losing the core IP identity.生成场景把每个角色扩展进小型叙事世界，同时不丢失核心 IP 识别。
02 — AIGC Process02 — AIGC 流程
AIGC mattered only when the output could be directed.AIGC 的价值在于可被控制地生成。
The case should be discussed through controllable decisions: role definition, emotional palette, material language, prompt iteration, reference control, and final selection. 这个项目适合从可控决策讲起：角色定义、情绪色彩、材质语言、提示词迭代、参考约束和最终筛选。
Emotion, role, material.情绪、身份、材质。
Each character starts from a mood and a role, then gets translated into outfit, setting, and object cues.每个角色从一种情绪和身份出发，再转译成服装、场景和物件线索。
Keep the toy language stable.保持玩偶语言稳定。
Proportion, glossy material, large eyes, soft textile details, and controlled lighting are used as visual anchors.比例、光泽材质、大眼睛、织物细节和受控光影，是保持一致性的锚点。
Select, compare, refine.筛选、比较、收敛。
The final value is not one lucky image, but a repeatable direction that can generate a family of related visuals.最终价值不是一张偶然好图，而是能持续生成同族视觉的方向。
03 — Visual Archive03 — 视觉归档
The pages are kept as a continuous visual appendix.这些页面作为连续视觉附录保留。
The original pages are shown without extra captions so the IP world, mood range, and generated-scene quality can be read continuously. 原始页面不额外加说明，让 IP 世界观、情绪范围和生成场景质量可以连续阅读。
