Scanned PDF to DOCX
Convert scanned PDFs that have no text layer (classical-Chinese works, old typeset academic books) into clean, well-structured Word (.docx) using MinerU OCR plus a three-tier proofreading workflow. Use when turning scanned book pages into editable Word, OCR-ing a PDF into .docx, or rebuilding a typeset book's chapters / sections / special symbols. Not for PDFs that already have a good text layer (extract those directly).
Install
npx skills add yanouyuan-bit/scanned-pdf-to-docx --skill scanned-pdf-to-docxgit clone --depth 1 https://github.com/yanouyuan-bit/scanned-pdf-to-docx.git /tmp/scanned-pdf-to-docx
mkdir -p ~/.claude/skills
cp -r /tmp/scanned-pdf-to-docx/. ~/.claude/skills/scanned-pdf-to-docxPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Scanned PDF → clean Word Pipeline: **split → OCR (MinerU) → title map → generate (convert.py) → three-tier proofread → deliver.** Scripts live in `scripts/`; a worked example is in `examples/yixue/`; a printable human guide is `docs/workflow-guide.docx`. > **Copyright:** the workflow and tools are open; the *book* is not. Never commit > or redistribute source PDFs, OCR text, or generated Word files. ## 0. When to use Use for scanned/image PDFs (no selectable text). If the PDF already has a clean text layer, skip OCR and extract directly — this skill is overkill. ## 1. Environment (once) MinerU needs Python 3.10–3.12 (not 3.13). Use an isolated env; first OCR run auto-downloads ~1 GB of models from HuggingFace (one time). ``` pip install -U uv uv venv C:\mineru-env --python 3.12 uv pip install --python C:\mineru-env\Scripts\python.exe -U "mineru[core]" python-docx pymupdf rapidocr-onnxruntime C:\mineru-env\Scripts\mineru.exe --version ``` Run every python command with that env's interpreter. `C:\mineru-env` is local (does not sync to OneDrive); a new machine needs a reinstall. ## 2. Split the PDF A whole book at once thrashes memory for hours; split into 50–400-page chapter parts and do one at a time (`fitz.open` → `insert_pdf(from_page, to_page)` → `save`). Render a couple of pages first (`render_pages.py`) and record the **printed-page → PDF-page offset** — you reuse it through the whole proofread. ## 3. OCR with MinerU
Details
- Skill name
scanned-pdf-to-docx- Repository
- yanouyuan-bit/scanned-pdf-to-docx
- License
- MIT
- Repo updated
- 18 July 2026
FAQ
What does the Scanned PDF to DOCX skill do?
Convert scanned PDFs that have no text layer (classical-Chinese works, old typeset academic books) into clean, well-structured Word (.docx) using MinerU OCR plus a three-tier proofreading workflow. Use when turning scanned book pages into editable Word, OCR-ing a PDF into .docx, or rebuilding a typeset book's chapters / sections / special symbols. Not for PDFs that already have a good text layer (extract those directly).
How do I install the Scanned PDF to DOCX skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add yanouyuan-bit/scanned-pdf-to-docx --skill scanned-pdf-to-docx. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Scanned PDF to DOCX popular?
It lives in yanouyuan-bit/scanned-pdf-to-docx, which has 19 GitHub stars and 1 forks.
Related skills
Heige Ppt
21HeiGeAi/HeiGe-PPT
演讲锻造系统。把每一页 PPT 讲成一句话，把整套 deck 导成一场演出，而不是把要点堆成一面字墙。 专门用于从零设计有气场、有记忆点、一眼不像 AI 做的演示文稿、汇报 deck、路演 pitch、产品 keynote。 适用场景：创业路演、产品发布、年度复盘、方案提案、技术分享、培训课件、述职汇报、招商招生。 核心方法论：一页一句话 / 黄金三秒 / 节奏编排 / 视觉锤 / 反 AI 体检。 当用户要求"做个 PPT""做套幻灯片""做个路演 deck""做个产品发布 keynote""把这份内容做成演示" "做个汇报""做个提案""把这个做成幻灯片""做个演讲稿配图"时，必须使用此 skill。 输出为单文件 HTML 幻灯片：16:9、键盘翻页、可投屏可导出 PDF，且默认带可编辑层，用户打开就能在浏览器里直接改幻灯片文字、再导出 PDF 或下载独立 HTML。动手前先定调，出货前过反 AI 体检。 需要给客户一份能在 PowerPoint / WPS 里直接改的 .pptx 时，用内置的 html2pptx 转换器把同一份 HTML 设计源转成可编辑 PPTX，一份设计两种格式，不用做两遍。 注意：本 skill 负责"从零设计有气场的演示 deck"；如果是做单张落地页/网页界面，那是 heige-ui 的活； 如果是把现成文档转成静态展示长页，那是 html-be
Resume Rewrite
18xiaowantree/data-engineer-resume-copilot
Rewrite, diagnose, and generate data engineering resume bullet points. Also matches resumes against job descriptions and generates interview prep questions. Use this skill when the user wants to improve, rewrite, or critique resume bullets, describe their work in plain language to generate bullets from scratch, compare their resume with a job posting, or prepare for interviews. Trigger on "帮我改简历", "帮我写简历", "rewrite my resume", "看看这个简历", "这条怎么改", "我不会写简历", "帮我看看这个JD", "我的简历match吗", "这个岗位合适吗", "帮我押题", "准备面试", "深挖简历", "interview prep", or any time the user pastes resume bullets, describes work ex
Resume Assistant
21Y1fe1-Yang/resume-assistant-skill
智能简历助手,通过五个AI代理提供全流程求职支持:(1)故事挖掘-发现经历亮点;(2)职位推荐-匹配合适岗位;(3)简历优化-针对JD定制内容;(4)模拟面试-实战演练与反馈;(5)能力提升-差距分析与计划。适用于简历创建、优化、面试准备、职业规划等求职相关任务。
Boss View
18yanliudesign/boss-view-skill
站在 CEO、CFO、CMO、COO、CPO、CTO、VP Design、VP Product 或直系老板视角，将零散工作记录生成结构化、突出业务价值、可导出 PDF/Markdown 的 Executive Update HTML 报告。适用于周报、工作汇报、Manager Sync、1:1。固定流程：接收输入→选择汇报对象→调用角色知识库→生成完整 8 章 HTML→补充证据。硬约束：绝不编造数据与结果，从"活动"转向"结果"。
Baoxiao
17xntj-ai/baoxiao
交互式生成合规的中国企业费用报销单 PDF 包。用户提供装有发票 PDF 和/或发票照片的文件夹,**先按统一规范整理发票文件名并按开票月归档(含照片)**,再通过提问(报销主体/报销人/部门/费用所属期/用途等)逐步生成:封面汇总单 + 分析合规页 + 逐票明细页(嵌发票图)+ 滴滴行程单紧凑重制 + 发票查重台账 CSV。当用户要求"做报销单""生成报销单""把这些发票整理成报销单""报销 PDF""费用报销"或提供一批发票要归集报销时使用。处理增值税电子发票(PDF,含文本层)和纸质发票照片(经 Claude 视觉识别)。
Ats Latex CV
19NoahMustafa/claude-ats-cv-skill
Build ATS-optimised, human-sounding CVs in LaTeX (Overleaf-ready code). Trigger this skill whenever a user asks to create, rewrite, reconstruct, or improve a CV, resume, or job application document — including when they upload an existing CV, paste raw experience, or describe a target role. Also trigger when the user says things like "make my CV better", "tailor my resume for X", "help me apply for a job", or "write my CV from scratch". Always use this skill for any CV/resume task, even if the user doesn't use the word "LaTeX" or "ATS".
Browse the most popular Claude skills, all documents & office files skills, or write your own with the skills generator.