
Academic Figure Prompt
This generates extremely detailed English prompts for AI image tools to create academic paper figures in the style of top-tier conferences like CVPR or NeurIPS. It walks you through understanding your paper content, analyzing reference figures if you have them, and then forces you to pick a color scheme from eight presets (Okabe-Ito, Navy+Coral, minimal grayscale, etc.) before writing anything. The prompts it creates are intentionally verbose and information-dense, covering every module, arrow, dimension label, and embedded thumbnail because the goal is precision over brevity. It's built for framework diagrams, architecture figures, and pipeline flowcharts where you need to pack in a lot of technical detail without looking like generic AI art. The color palette enforcement is smart since that's usually what separates a real paper figure from something that screams "made with Midjourney."
- GitHub stars
- 2K
- last updated
- Aug 21, 2026
Install
$ npx -y skills add ligphidonk/academic-figure-generator --skill 'Academic Figure Prompt' --agent claude-codeInstalls into .claude/skills of the current project.
Files
Academic Figure Prompt — 学术论文配图提示词生成器
为学术论文生成极其详细的英文提示词,供 AI 图片生成工具(NanoBanana / Gemini / Midjourney / DALL-E)生成顶会级别的专业学术配图。
核心理念
生成的提示词必须做到三点:信息密度极高、视觉风格精确、内容完整无遗漏。
宁可提示词过长过详细,也绝不能简化省略。学术配图的价值在于精准传达复杂信息,而非美观简洁。
工作流程
Step 1: 理解论文内容
在生成提示词之前,必须先充分理解论文内容:
- 阅读用户提供的论文/章节源文件(LaTeX、Word、PDF 等)
- 提取每个章节的核心概念、方法、模型架构、数据流
- 识别所有需要配图的位置及其内容需求
- 理解论文中的数学符号、变量含义、维度信息
Step 2: 分析参考图(如有)
如果用户提供了参考图,必须详细分析:
| 分析维度 | 提取内容 |
|---|---|
| 配色方案 | 主色、辅色、强调色的精确色值(如 #3A8F85) |
| 布局结构 | 流向(左→右 / 上→下)、分区方式、层次关系 |
| 模块样式 | 框的形状、边框粗细、填充色、圆角大小 |
| 标注方式 | 标题栏样式、公式标注、维度标注、箭头类型 |
| 信息密度 | 每个模块内的子细节数量、嵌入缩略图的使用方式 |
| 特殊元素 | 反馈环路、虚线框、跳接箭头、图例位置 |
Step 2.5: 配色方案选择(必须在生成提示词前完成)
在生成任何提示词之前,必须先向用户展示配色选项并等待确认。
展示以下内容:
请选择配色方案(输入编号或自定义):
Sponsors
Details
- Repository
- github.com/ligphidonk/academic-figure-generator
- Categories
- DevOps & CI/CD, Code Review & Quality, AI & Agent Building, Automation & Workflows, Sales & Marketing, Design & UI/UX, Office & Documents, Generative Media
- First seen
- Jun 3, 2026