# paper2anything [![Project Page](https://img.shields.io/badge/Project_Page-Live-2ea44f?logo=githubpages&logoColor=white)](https://quzhan51496.github.io/paper2anything/) [![Claude Code Skill](https://img.shields.io/badge/Claude_Code-Skill-blue?logo=anthropic&logoColor=white)](https://docs.anthropic.com/en/docs/claude-code) [![License: Apache 2.0](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](LICENSE) **English** | [简体中文](README.zh-CN.md) [Showcase](#showcase) · [Installation](#installation) · [Usage](#usage) · [Output locations](#output-locations) · [Directory structure](#directory-structure) · [Contributors](#contributors) · [Acknowledgements](#acknowledgements)

A marquee of real artifacts produced by paper2anything — slide decks, posters, project homepages, Xiaohongshu and WeChat covers

A skills pack that turns an academic paper PDF into any of several promotional artifacts, covering the main forms of paper dissemination: | Skill | Turns the paper into | Output | Trigger phrases (examples) | | --- | --- | --- | --- | | **paper2slides** | a presentation deck | `.pptx` | "make slides from this paper", "generate a deck from this PDF", "deck this paper" | | **paper2poster** | a conference poster | `poster.html` + `poster.png` | "make a poster from this paper", "turn this paper into a poster" | | **paper2html** | a single-page project homepage | `index.html` | "turn this paper into a webpage", "generate a project page / landing page" | | **paper2xhs** | a Xiaohongshu (rednote) multi-image post | `xhs_post.json/md` + cover + figure images | "post this paper to Xiaohongshu", "turn this paper into a rednote post" | | **paper2wechat** | a WeChat Official Account article | `wechat_article.md/html` + cover | "write this paper as a WeChat article", "turn this paper into a WeChat post" | Each subdirectory is an independent, auto-triggerable skill (each has its own `SKILL.md`). --- ## Showcase ### paper2slides A sample of **paper2slides** output — eight decks across diverse research fields (HCI / VR, computer vision, drug discovery, security & provenance, multimodal agents, quantum computing, optimization, and ML theory):

Eight slide decks generated by paper2slides from paper PDFs, across diverse research fields

### paper2poster A sample of **paper2poster** output — nine posters, spanning nine research fields (drug design, graphics, HCI, materials science, X-ray inspection, NLP / LLM safety, optics & photonics, statistical ML, and computer vision):

Nine conference posters generated by paper2poster from paper PDFs, across nine research fields

### paper2html A sample of **paper2html** output — eight project homepages across diverse research fields (HCI / VR, drug discovery, machine writing, robotics, security, software engineering, agent skills, and financial trading):

Eight project homepages generated by paper2html from paper PDFs, across diverse research fields

### paper2xhs A sample of **paper2xhs** output — ten Xiaohongshu (rednote) posts across diverse research fields (world models, LLM agent societies, AI supply-chain security, LLM evaluation, robotic manipulation, CS education, machine translation, climate simulation, medical diagnosis, and causal inference):

Ten Xiaohongshu posts generated by paper2xhs from paper PDFs, across diverse research fields

### paper2wechat A sample of **paper2wechat** output — twelve WeChat Official Account articles across diverse research fields (world models, LLM agent societies, AI supply-chain security, LLM evaluation, robotic manipulation, CS education, machine translation, climate simulation, medical diagnosis, causal inference, efficient inference, and NLP):

Twelve WeChat Official Account covers generated by paper2wechat from paper PDFs, across diverse research fields

--- ## Installation **One-command install with Claude Code (recommended)**: ```bash bash tools/install-linux.sh --create-env --shell-init # Linux bash tools/install-macos.sh --create-env --shell-init # macOS ``` The install script will: ① symlink the 5 skills into `~/.claude/skills/`; ② bootstrap a `.env` from `.env.example` if you don't have one; ③ check the conda environment, system dependencies, and `MINERU_API_TOKEN` (used for paper extraction). What the two flags do (omit them if you don't need them): - `--create-env`: `conda env create` (updates per `environment.yml` if it already exists) + installs playwright chromium + runs a pip self-check; - `--shell-init`: (optional) writes the `.env` auto-export into your shell startup file, so a new shell loads the credentials automatically. After it runs, follow the script's prompts for two more steps: 1. **Fill in `MINERU_API_TOKEN` in `.env`** (required); 2. **Install the missing system-level dependencies it points out** (the script detects each one and gives the install command). If you need to install manually (e.g. you only use one specific skill), here is the equivalent breakdown.

Manual install

### Register a skill Symlink the skills you want into `~/.claude/skills/` so Claude Code can discover and auto-trigger them: ```bash # From the paper2anything package root; copy the line for whichever skill you want mkdir -p ~/.claude/skills ln -sfn "$(pwd)/paper2slides" ~/.claude/skills/paper2slides ln -sfn "$(pwd)/paper2poster" ~/.claude/skills/paper2poster ln -sfn "$(pwd)/paper2html" ~/.claude/skills/paper2html ln -sfn "$(pwd)/paper2xhs" ~/.claude/skills/paper2xhs ln -sfn "$(pwd)/paper2wechat" ~/.claude/skills/paper2wechat ``` ### conda environment All 5 skills share one conda environment, `paper2anything`. ```bash # From the paper2anything package root conda env create -f environment.yml conda activate paper2anything ``` ### System-level dependencies | Tool | Purpose | Which skill | Install command | | --- | --- | --- | --- | | poppler-utils (`pdftoppm`) | PDF rendering | paper2slides | `sudo apt install poppler-utils` (Linux) / `brew install poppler` (macOS) | | libreoffice (`soffice`) | visual QA | paper2slides | `sudo apt install libreoffice` (Linux) / `brew install --cask libreoffice` (macOS) | | Node.js | JS runtime | paper2slides | on Linux use NodeSource (the apt default is too old, see the note below) / `brew install node` (macOS) | | pptxgenjs + react-icons/react/react-dom/sharp | PPT rendering | paper2slides | `npm install -g pptxgenjs react-icons react react-dom sharp` (prefix `sudo` on Linux) | **Node.js (Linux)**: sharp requires Node ≥20.9.0; install it via NodeSource: ```bash curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash - && sudo apt install -y nodejs ``` ### Credentials All skills' credentials live in **a single `.env` at the package root** (copy from `.env.example` and fill it in). After copying, just fill in your keys: ```bash cp .env.example .env # first time: copy, then fill in your keys ``` Optional: write the `.env` export into your shell startup file. Pass `--shell-init` during install to do this automatically (see Installation), or add this line manually: ```bash set -a; source /.env; set +a ```

--- ## Usage Just state your intent and the matching skill auto-triggers, e.g.: - "make `path/to/paper.pdf` into slides" → **paper2slides** - "make a conference poster from this paper" → **paper2poster** - "turn this PDF into a project homepage" → **paper2html** - "post this paper to Xiaohongshu for me" → **paper2xhs** - "write this paper as a WeChat article" → **paper2wechat** Or call them explicitly with a slash command, followed by the PDF path: ``` /paper2slides path/to/paper.pdf /paper2poster path/to/paper.pdf /paper2html path/to/paper.pdf /paper2xhs path/to/paper.pdf /paper2wechat path/to/paper.pdf ``` --- ## Output locations Each skill's final deliverable lands in the same directory as the paper; all intermediate artifacts are kept under `.paper2anything///` in that same directory: | Skill | Final deliverable | Intermediate artifacts | | --- | --- | --- | | paper2slides | `_slides/` (`.pptx`) | `.paper2anything/slides//` | | paper2poster | `_poster/` (`poster.png` + `poster.html` + `images/`) | `.paper2anything/poster//` | | paper2html | `_html/` (`index.html` + `images/`) | `.paper2anything/html//` | | paper2xhs | `_xhs/` (`xhs_post.md` + `.json` + `cover.png` + `post_images/`) | `.paper2anything/xhs//` | | paper2wechat | `_wechat/` (`wechat_article.md` + `.json` + `cover.jpg` + `figures/`) | `.paper2anything/wechat//` | --- ## Directory structure ``` paper2anything/ ├── environment.yml # python environment ├── .env.example # credentials template (copy to .env and fill in) ├── .gitignore # ignores .env / __pycache__ etc. ├── LICENSE # Apache-2.0 ├── README.md # this file (English, default) ├── README.zh-CN.md # Chinese version ├── tools/ # install scripts ├── assets/ # static files │ └── showcase/ # per-skill output samples (assets/showcase//) ├── paper2slides/ # paper → slides │ ├── SKILL.md │ ├── references/ # design style, outline heuristics, pipeline, schema, pptxgenjs │ └── scripts/ # parse_pdf / render_pptx / page_screenshot / workdir + lib/ ├── paper2poster/ # paper → poster HTML/PNG │ ├── SKILL.md │ ├── references/ # poster examples, color palettes, layout guide │ └── scripts/ # parse_pdf / auto_outline / geom_check / collect_figures / screenshot / check_env ├── paper2html/ # paper → single-page project homepage │ ├── SKILL.md │ ├── references/ # design languages, HTML authoring spec, QA checklist │ └── scripts/ # parse_pdf / validate / render_check etc. + lib/ (parse/extract/QA, no renderer) ├── paper2xhs/ # paper → Xiaohongshu │ ├── SKILL.md │ ├── references/ # publish guide │ └── scripts/ # parse_pdf / cover / post_images / publish / xhs_login + utils └── paper2wechat/ # paper → WeChat ├── SKILL.md └── scripts/ # parse_pdf / cover / publish + utils ``` --- ## Contributors Skill design leads at the **[AI4GC Lab](https://ai4gc.org/)**:

_{Honghui Sheng}
_paper2poster

_{Tao Xiong}
_paper2html

_{Xinchen Xu}
_{paper2xhs & paper2wechat}

--- ## Acknowledgements paper2anything is developed by the **[AI4GC Lab](https://ai4gc.org/)** at Zhejiang University. - Paper PDF parsing is powered by **[MinerU](https://github.com/opendatalab/MinerU)**. - Xiaohongshu publishing is powered by **[xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp)**. - WeChat publishing and formatting are powered by **[md2wechat](https://pypi.org/project/md2wechat/)**.