Paper Explainer logo

Paper Explainer

OrganizationPopular
ZJU-REAL
paper-explainer

科研论文解读:解析 arXiv/PDF 的公式与图表,提炼问题、贡献、方法、关键图和结论,再产出 B站/视频号解读视频或知乎/公众号图文。 当用户说“论文解读、讲论文、论文转视频/图文、科研科普、arXiv、学术视频”时使用。 本 SKILL 从论文做内容;video-to-article 从视频做图文,doc-convert 只转换文档格式。

Overview

PublisherZJU-REAL
RepositoryEasel
Skill namepaper-explainer
Stars
1.2K
Forks
175
Bundled files
10
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 10 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by ZJU-REAL on GitHub. Read the source before you install it.

Installation

Install the Paper Explainer AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/ZJU-REAL/Easel.git /tmp/Easel
mkdir -p .claude/skills
cp -r /tmp/Easel/skills/openclaw/paper-explainer .claude/skills/paper-explainer
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Paper Explainer in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Paper Explainer on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Paper Explainer is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

科研论文解读(论文 → 视频 / 图文)

把一篇论文讲成普通人/同行都爱看的视频号视频或图文。核心中间产物是一份 结构化 asset library(一次解析+提炼,视频与图文两条产线共用,不重复调 LLM)。 确定性 IO(拉论文/解析 PDF/骨架)走 scripts/paper_ingest.py提炼与分镜脚本由你 LLM 完成——这是本 SKILL 的核心价值。

视频转图文(反向)见 video-to-article;纯格式转换见 doc-convert; 只做图表见 chart-visualization / infographic;发视频号见 skill-channels-upload

输入

字段必填说明
论文arxiv id(2401.12345)/ arxiv 链接 / 本地 PDF 路径(没给就问)
目标形态视频(默认,视频号/B站)/ 图文(知乎/公众号)/ 两者都要
视频画幅视频时必填用户或上游任务未明确横版/竖版(或 16:9/9:16/具体分辨率)时,进入视频制作前必须追问并等确认;不得按平台、Profile 或默认值静默推断,已明确则不重复问
受众深度大众科普(默认)/ 同行向(更专业)
时长视频默认 2–4 分钟(视频号中视频)

产物结构(outputs/论文简称/

article.md               图文版(知乎/公众号)
final.mp4                成片
assets/                  paper.pdf / parsed/ / asset-library.json / script.md
  slide-plan.json        结构化分页(页面唯一输入,口播与屏幕文字分离)
  slides/                稳定渲染的逐页 PNG + HTML + audit report
  slides-contact-sheet.jpg  整套视觉复核图

脚本(相对项目根):paper_ingest.py(解析)+ render_slides.py(分页校验/渲染/审计)。

执行步骤

1. 取原文 + 解析

  1. 环境自检python skills/openclaw/paper-explainer/scripts/paper_ingest.py check (看 pdfplumber / MinerU token / 代理;缺 pdfplumber 则 pip install pdfplumber)。
  2. 拉论文paper_ingest.py fetch --paper <id/url/本地pdf> -o outputs/论文简称/assets/paper.pdf
  3. 解析paper_ingest.py parse -i outputs/论文简称/assets/paper.pdf -o outputs/论文简称/assets/parsed/ (有 MINERU_API_TOKEN 走 MinerU 含公式/图表结构化,否则 pdfplumber 纯文本 + 尽力抽图)。

2. 结构化提炼(你来做,核心)

  1. 生成骨架:paper_ingest.py skeleton -o outputs/论文简称/assets/asset-library.json
  2. assets/parsed/content.*,按 references/paper-distill-schema.md 填满 assets/asset-library.jsonone_liner(一句话讲清干了啥)、problem/prior_gapcontributions(≤3 条)、 method(含通俗类比 analogy)、key_figures(挑 2–4 张关键图,每张写 plain 大白话解释)、 results(含关键数字)、limitationstakeawayterms(术语通俗表)。 通俗化方法references/explain-methodology.md(公式/图表→大白话、类比法、避免堆术语)。
  3. 忠于原文:不夸大、不编造结论;拿不准的地方标注,别臆测(学术内容错了会被同行抓)。

3A. 视频产线(视频号/B站)

  1. 分镜脚本:按 references/video-storyboard.md 结构(钩子→问题→已有不足→贡献→方法一图讲清→结果→意义)把 asset-library 写成 assets/script.md。 分镜/留存/口播节奏复用 video-script 的方法(喂论文语境)。可选双人问答口播(主持人提问+讲解者回答)比单人旁白更抓耳——用双人时把口播写成逐行 lines.json{speaker,text,emotion},speaker=主讲/提问)。
  2. 视觉素材盘点 + 配图:先列出每页的视觉角色(证据图/重绘图/概念线稿/字体图形/motif),再写 slide-plan。论文原图从 assets/parsed/figures/ 选用;复杂原图先裁关键区域,方法流程/结果图用 infographic / chart-visualization 重绘。封面/概念页缺图时,主动找或制作与主题直接相关的线稿、局部图或符号素材,不用随机机器人/blob 填空。图中文字在目标分辨率不可读就不得直接使用。
  3. 稳定 slide 产线(强制,不得在 outputs 临时写 make_slides 脚本):先读 card-design 的设计原则和 references/slide-design.md,把 script 写成 assets/slide-plan.json。把用户/Profile/参考图的原始风格意图原样写入 style,再分别选 base_styletreatmentthememotif 和视觉素材来实现;不得把用户风格强行归为某个预设,也不得因没有同名预设而拒绝。迁移的是可观察特征(氛围、配色、线条、纹理、构图、角色/物件素材),不是穷举风格名。未指定风格时用 editorial,但默认也必须有明确的编辑网格、纸张层次、章节锚点和图片框法,不得交付“素底 + 字”。整套锁定一个设计立场,页面骨架与审计门保持稳定。运行:
    bash
    python skills/openclaw/paper-explainer/scripts/render_slides.py validate --plan outputs/<项目>/assets/slide-plan.json
    python skills/openclaw/paper-explainer/scripts/render_slides.py render --plan outputs/<项目>/assets/slide-plan.json --out-dir outputs/<项目>/assets/slides
    python skills/openclaw/paper-explainer/scripts/render_slides.py audit --plan outputs/<项目>/assets/slide-plan.json --slides-dir outputs/<项目>/assets/slides --contact-sheet outputs/<项目>/assets/slides-contact-sheet.jpg
    任一非 0 退出必须改 plan 后重渲;validate 会按页面职能拦截“只有口号、缺少解释”的低信息页,并检查合并主题后所有正文色在实际背景上的对比度;明亮 accent 可继续用于装饰,文字会使用可读的语义前景色。render 会硬拦文字/元素越界、重叠、组内不对齐、卡内文字左边漂移与结构页过度空洞。脚本全过后当前 Agent 必须肉眼查看 contact sheet 和至少 3 张原尺寸 slide,检查暂停/静音时页面能否独立读懂、论文图可读、文字是否和所属元素对齐、留白是否有叙事作用、视觉素材是否相关、节奏是否重复;只过脚本不等于合格。不要把 narration 整段搬上屏。只有论文图本身承载主要信息时才可在该页设 density: visual,不得把它当作跳过内容提炼的开关。
  4. 成片(配音+字幕+合成,缺一不可):从 slide-plan 的 narration 生成口播——单人用 tts-voiceover,双人用 multi-voice-dubbing;同步 SRT,缺则跑 auto-subtitle。把 assets/slides/slide_*.png、配音和字幕写入 auto-short-video storyboard 后合成,必须设顶层 "image_motion": "static";slide/图表禁用 Ken Burns,不得缩放、平移或裁掉边缘。页面停留时长按对应 narration 音频/字幕分段,不均分整轨。不能只交静态图或无声视频。
  5. manifest.py meta 登记 final.mp4article.md 为 deliverable;中间解析、slide 和音频只放 assets/
  6. 发布:交 skill-channels-upload(视频号)/ B站 biliup。

3B. 图文产线(知乎/公众号)

  1. 同一份 asset-libraryarticle.md:标题(钩子)+ 用大白话讲清 problem→method→results→takeaway,配 assets/ 的图。 平台适配见 references/platform-adapt.md(知乎逻辑链、公众号成文起承转合)。排版/长图交 doc-convert;发布交 skill-zhihu-publisher / skill-wechat-publisher

Profile 感知

  • 有 Profileplatforms.md 定主平台并给出形态/画幅/时长建议,但视频画幅仍须用户确认;audience.md 定受众深度(大众 vs 同行);style.md 定讲解调性;identity.md 定领域垂类(AI/生物/材料…影响类比取材)。
  • 无 Profile:默认视频号 2–4 分钟中视频、大众科普深度,先问领域与受众。

规则

  1. 忠于原文:不夸大贡献、不编造数字/结论;术语拿不准先查原文,别臆测。
  2. 一次提炼、两处复用:asset-library.json 是唯一真相源,视频与图文都从它出,避免重复提炼与口径不一。
  3. 通俗但不失真:用类比降低门槛,但类比不能扭曲原意;关键术语给一句通俗解释而非回避。
  4. 图优先:论文靠图讲方法/结果,视频/图文尽量用图(原图或重绘信息图)承载信息。
  5. 刻意不做:数字人讲座(太重)、依赖 LaTeX 源(从 PDF 入覆盖更广)。
  6. 页面不是口播稿,也不是口号板:一页一个中心结论,但必须用解释、证据或数字口径让页面在暂停/静音时也能独立读懂;细节留给 narration,不得靠缩小字号容纳过量文字。

参考来源

EASEL-META.md。流程沉淀自 QuZhan51496/paper2anything(本身即 Claude Skills:parse_pdf/MinerU + 提炼方法论外置 references + 多形态扇出)、 showlab/Paper2Video(按内容块切段、字幕先行)、Paper2Poster(结构化 asset library 中间产物)、 Azzedde/paper_to_podcast(双人问答口播)、OpenDCAI/Paper2Any(一次解析扇出多形态)。

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Paper Explainer AI skill do?

科研论文解读:解析 arXiv/PDF 的公式与图表,提炼问题、贡献、方法、关键图和结论,再产出 B站/视频号解读视频或知乎/公众号图文。 当用户说“论文解读、讲论文、论文转视频/图文、科研科普、arXiv、学术视频”时使用。 本 SKILL 从论文做内容;video-to-article 从视频做图文,doc-convert 只转换文档格式。

Why use Paper Explainer on TypingMind?

Because you install it once and use it with any model. Paper Explainer is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Paper Explainer in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/ZJU-REAL/Easel/tree/main/skills/openclaw/paper-explainer. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Paper Explainer?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Paper Explainer?

As many as you like. As long as a model supports skills, you can use Paper Explainer with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Paper Explainer AI skill free?

Yes. It is published on GitHub by ZJU-REAL under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇