Paper2slides

Generate presentation-ready slides from research PDFs and related documents with LLM + RAG.

Paper2slides is a local workflow for turning papers, reports, and related source files into academic presentation decks. It parses PDF and document content, extracts figures and formulas, plans slide structure with an LLM, and exports both Beamer and native PPTX outputs.

Features

• Parse PDF, DOCX, and TXT inputs
• Merge multiple source files into one workflow
• Extract text, figures, tables, and formulas from source documents
• Plan slide outlines and per-slide content with LLM assistance
• Export Beamer LaTeX, compiled PDF, PPTX, and optional slide preview images
• Support both CLI and Streamlit web UI
• Keep logs and intermediate outputs in local project folders

Typical Use Cases

• Paper presentation slides
• Thesis defense decks
• Research group report slides
• Project pitch and internal technical briefings

Quick Start

1. Install dependencies

pip install -r requirements.txt

2. Configure model access

Create a .env file in the project root:

APIYI_BASE_URL=your_openai_compatible_base_url
APIYI_API_KEY=your_api_key
LLM_MODEL=deepseek-chat

3. Start the web UI

streamlit run streamlit_app.py

4. Or run the CLI

python run_agent.py

Outputs

Paper2slides can generate:

• .tex Beamer source
• .pdf compiled slide deck
• .pptx native PowerPoint file
• optional exported slide preview images

Project Structure

.
|- streamlit_app.py
|- run_agent.py
|- ppt_agent/
|- tests/
|- figures/
`- output/

Tech Stack

• Python
• Streamlit
• LangChain / LangGraph
• OpenAI-compatible APIs
• PyMuPDF / pdfplumber / pypdf
• python-pptx / python-docx / reportlab

Status

The current repository already includes the core parsing, planning, export, and local UI workflow. It is suitable as a working local tool and can be extended further with custom templates, prompt strategies, and output styles.