Built an offline-first AI tool that deeply explains research papers instead of summarizing them away.
Drop in a PDF or arXiv link and get a structured walkthrough with narration, synced highlights, and a single polished video output.
Designed for actually understanding papers — not skimming them.
What it does
- Produces a high-level overview and structured explanations.
- Walks through the paper paragraph by paragraph with narration.
- Visually highlights the current paragraph as it’s being explained.
- Outputs a single MP4 combining visuals and audio.
- Works fully offline, with optional Gemini support for higher-quality reasoning.
- Modular pipeline that’s easy to extend and experiment with.
How it works
- Accepts PDFs from local paths or URLs.
- Parses and structures paper content for sequential explanation.
- Uses Gemini with an Ollama fallback for robustness.
- Custom TTS, highlighting, and video composition pipeline.
- Intermediate artifacts are stored for reproducibility and iteration.