🌐 中文文档 | English

📡 PaperCast — One Paper, Every Format

Drop a paper. Get everything you need to promote it.

PaperCast reads your research paper (PDF or arXiv link) and auto-generates:

• 🐦 X/Twitter Thread — 5-8 tweets, hook-first
• 💼 LinkedIn Post — professional long-form
• 🇨🇳 Chinese Summary — 公众号/知乎 ready
• 🎨 Graphical Abstract — AI-generated visual summary
• 🎤 Talk Outline — 10-min presentation structure
• 📰 Press Release — journalist-ready narrative

---

Why

Academics are terrible at promoting their own work:

Problem	PaperCast
Writing social posts takes hours	30 seconds
Non-English audiences unreachable	Auto Chinese summary
No design skills for graphics	AI graphical abstract
Press releases need PR training	Auto-generated
Every paper, same format inconsistency	Consistent brand voice

Inspired by Typefully — the best social media writing tool for creators. PaperCast is Typefully for academics.

---

How It Works

PDF / arXiv / Title+Abstract
        │
        ▼
   LLM reads & extracts
   - Core finding
   - Key numbers
   - So-what
        │
        ▼
   Format templates
   ┌──────┼──────┬──────┬──────┐
   ▼      ▼      ▼      ▼      ▼
 Thread LinkedIn Chinese Graph Press

---

MVP — Phase 1

Web UI (Streamlit). Input: PDF upload or arXiv URL. Output: 3 formats.

Format	Status
X/Twitter Thread	✅ MVP
LinkedIn Post	✅ MVP
Chinese Summary	✅ MVP
Graphical Abstract	🔜 v0.2
Talk Outline	🔜 v0.2
Press Release	🔜 v0.2

---

Quick Start

git clone https://github.com/SpencerRaw/papercast.git
cd papercast
pip install -r requirements.txt
streamlit run app/streamlit_app.py

Set your API key:

export DEEPSEEK_API_KEY=sk-...
# or
export OPENROUTER_API_KEY=sk-or-...

---

Comparison

	Typefully	PaperCast
Domain	Social media creators	Academic researchers
Input	Ideas / drafts	Papers (PDF, arXiv)
Output	Tweets, LinkedIn posts	Threads, LinkedIn, Chinese, Graphics, Press
AI depth	Rewrite/suggest	Read paper → extract → format
Target user	Anyone building audience	Researchers who want their work seen

---

Tech Stack

• LLM: DeepSeek / OpenRouter (paper comprehension + generation)
• PDF: PyMuPDF (text extraction)
• Frontend: Streamlit
• Prompt chain: Extract → Hook generation → Format-specific rendering

---

License

MIT