SLIDE SUMMARIZER

A web application that converts PowerPoint presentations into detailed text documents using AI vision models.

Purpose

Slide Summarizer solves the challenge of converting PowerPoint presentations into text-based formats that can be effectively processed by language models and other text-based systems. While PowerPoint is excellent for visual storytelling and human audiences, its content is difficult to process through traditional text scraping methods because:

• Slides often contain minimal text, relying on visual elements to convey information
• Charts, graphs, and visual elements lose their semantic meaning when converted to plain text
• Traditional text extraction misses the rich context and relationships between visual elements
• The format is not ideal for retrieval augmented generation (RAG) systems that primarily work with text

PPT2Doc bridges this gap by using AI vision models to:

• Accurately transcribe all text content from slides
• Provide detailed descriptions of visual elements like charts and graphs
• Preserve the semantic relationships between different elements
• Generate text that maintains the original meaning and context

Pipeline

1. Input: User uploads a PowerPoint (.pptx) file
2. Processing:
   • PowerPoint is converted to PDF using LibreOffice
   • PDF is converted to individual PNG images (one per slide)
   • Each slide image is processed by AI vision model (OpenAI o4-mini or Google Gemini)
   • AI model transcribes text and describes visual elements
3. Output: Generated text document containing transcribed content from all slides

Tech Stack

Frontend

• React with TypeScript
• Tailwind CSS for styling
• Vite for build tooling

Backend

• FastAPI (Python)
• OpenAI API (o4-mini model)
• Google Gemini API
• LibreOffice for PPTX to PDF conversion
• pdf2image for PDF to PNG conversion
• Docker for containerization

Features

• Upload PowerPoint (.ppt, .pptx) files
• Convert slides to text using OCR
• Generate AI-powered summaries using either:
  • OpenAI GPT-3.5 Turbo
  • Google Gemini AI
• Real-time processing status updates
• Modern, responsive UI
• Support for both concise and detailed summaries