paper-comic

Original🇨🇳 Chinese
Translated

Generate educational comics from academic papers, explaining core viewpoints and innovations through visual storytelling. Supports 4 art styles: classic (clean line), tech (technical), warm (warm tone), chalk (chalkboard). Uses baoyu-gemini-web to generate images.

9installs
Added on

NPX Install

npx skill4agent add zsyggg/paper-craft-skills paper-comic

SKILL.md Content (Chinese)

View Translation Comparison →

Paper Comic - Academic Paper Comic Generator

Convert academic papers into coherent educational comics, making complex concepts easy to understand through visual storytelling.

Usage

bash
/paper-comic /path/to/paper.pdf
/paper-comic /path/to/paper.pdf --style tech
/paper-comic  # Then paste the paper content

Art Style Options

Art StyleFeaturesSuitable Papers
classicClean line style, concise and professional, easy to understandGeneral purpose, suitable for most papers (default)
techTechnical style with circuit elements and neon effectsAI/Computer Science/Engineering papers
warmWarm style, approachable, nostalgicPsychology/Cognitive Science/Education papers
chalkChalkboard style, blackboard effect, academic feelMathematics/Physics/Theoretical papers

Output Structure

[output-dir]/
├── outline.md           # Outline and storyboards
├── characters/
│   ├── characters.md    # Character definitions
│   └── characters.png   # Character reference images
├── prompts/
│   ├── 00-cover.md      # Cover prompt
│   └── XX-page.md       # Page-specific prompts
├── 00-cover.png         # Cover image
└── XX-page.png          # Comic pages
Output Directories
  • With source files:
    [source-dir]/comic/
  • Without source files:
    comic-outputs/YYYY-MM-DD/[topic-slug]/

Workflow

Step 1: Analyze the Paper

  1. Read paper content (PDF or Markdown)
  2. Extract core information:
    • Paper title, authors
    • Research background and motivation
    • Core innovations (1-3)
    • Key methods/algorithms
    • Main experimental results
  3. Automatically recommend an art style based on the paper's field (or use user-specified style)

Step 2: Design Narrative Structure

Four-stage structure (suitable for 8-12 pages):
StageNumber of PagesContent
Introduction1-2 pagesProblem background, why this research is needed
Exploration2-3 pagesLimitations of existing methods, leading to innovations
Core3-5 pagesDetailed explanation of innovative methods, visualized with metaphors
Summary1-2 pagesExperimental results, significance and future outlook

Step 3: Define Characters

Create
characters/characters.md
:
Required Characters:
  • Mentor: Knowledge explainer, wise and approachable
  • Student: Represents readers, asks questions and learns
  • Concept Personification (optional): Personify abstract concepts
Character Consistency Requirements:
  • Mentor and student must appear in ≥60% of pages
  • Clearly mark characters present on each page
  • Character appearance remains consistent across all pages

Step 4: Create Storyboards

Create
outline.md
, including:
  • Metadata (title, art style, number of pages)
  • Cover design
  • Panel layout and content for each page
Storyboard Requirements:
  • 3-5 panels per page
  • Mark characters, scenes, and dialogue for each panel
  • All dialogue must be in Chinese
  • Express formulas with illustrations instead of text

Step 5: Generate Images

Generate using baoyu-gemini-web (requires Google account authentication):
bash
# Get skill installation path (assuming installed via npx skills add)
SKILL_DIR="$HOME/.claude/skills/baoyu-gemini-web"
# Or if located elsewhere:
# SKILL_DIR="$HOME/.codex/skills/baoyu-gemini-web"

# Generate character reference images
npx -y bun "$SKILL_DIR/scripts/main.ts" \
  --promptfiles references/base-prompt.md characters/characters.md \
  --image characters/characters.png \
  --sessionId comic-[topic]-[timestamp]

# Generate each page (use the same sessionId for consistency)
npx -y bun "$SKILL_DIR/scripts/main.ts" \
  --promptfiles references/base-prompt.md prompts/XX-page.md \
  --image XX-page.png \
  --sessionId comic-[topic]-[timestamp]
Key: Use the same
--sessionId
to ensure consistent character appearance.
First Run: Chrome browser will open for Google account authentication, and cookies will be cached afterward.

Step 6: Output Documentation

Generate
[topic]-paper-comic.md
:
markdown
# [Paper Title] - Comic Interpretation

## Overview
- **Paper**: [Title]
- **Art Style**: [Selected Style]
- **Number of Pages**: [N]
- **Generation Time**: [YYYY-MM-DD]

## Comic Pages

### Cover
![Cover](00-cover.png)

### Page 1
![Page 1](01-page.png)
**Content**: [Brief description of what this page explains]

...

## Core Knowledge Points
1. [Knowledge Point 1]
2. [Knowledge Point 2]
3. [Knowledge Point 3]

Key Principles

Text Requirements

  • All dialogue and narration must be in Chinese
  • Professional terms: Chinese + English, e.g., "梯度下降 (Gradient Descent)"
  • Text must be clear and readable, no blurriness

Formula Handling

  • Do not write formulas as text
  • Express formula meanings with illustrations/metaphors
  • Example: Gradient Descent → Draw a ball rolling down a hillside

Visual Coherence

  • Consistent character appearance throughout
  • Unified scene style
  • Clear and progressive narrative logic

Reference Files

  • references/base-prompt.md
    - Base prompt template
  • references/styles/classic.md
    - Clean line style template
  • references/styles/tech.md
    - Technical style template
  • references/styles/warm.md
    - Warm style template
  • references/styles/chalk.md
    - Chalkboard style template