steipete/summarize

What it solves

Summarize is a tool designed to extract clean, readable text from a wide variety of sources—including web pages, PDFs, images, and audio/video files—and generate concise summaries using Large Language Models (LLMs). It eliminates the noise of web pages and removes the need to manually transcribe media before summarizing it.

How it works

The tool operates through a Node.js CLI, a Chrome side panel, and a Firefox sidebar. It follows a multi-step processing path based on the input type:

• Web Content: Uses readability extraction with optional Markdown and Firecrawl fallbacks to get clean text.
• Media: Prioritizes existing captions and transcripts; if unavailable, it uses local or cloud transcription (including ONNX transcription) and tools like ffmpeg and yt-dlp for processing.
• Visuals: Extracts scene-based frames from videos and can perform OCR to turn images or slides into text.
• LLM Integration: Connects to various providers via API, OpenAI-compatible endpoints, Ollama, or authenticated coding CLIs (like GitHub Copilot or Claude) to generate the final summary.

Who it’s for

It is built for researchers, students, and professionals who need to quickly digest information from diverse media formats (web, video, podcasts, and documents) across different platforms.

Highlights

• Multi-source support: Handles web pages, PDFs, images, YouTube videos, and podcasts.
• Cross-platform access: Available as a CLI, Chrome extension, and Firefox sidebar.
• Advanced media processing: Supports local transcription, OCR for video slides, and timestamped output.
• Flexible model selection: Works with a wide range of LLM providers and local models via Ollama.
• Multilingual support: Interface and summary generation support numerous global languages.