Video To Text

video-to-text

[](https://opensource.org/licenses/MIT) [](https://github.com/adhenawer/video-to-text/releases) [](https://adhenawer.net/) [](https://www.python.org/) [](https://github.com/ml-explore/mlx) [](https://workers.cloudflare.com/)

🇧🇷 **[Leia em português brasileiro](README-pt_br.md)** · 🇺🇸 You are reading in English

Turns YouTube and Twitter/X videos and podcasts into readable posts — organized by sections and published as static HTML.

---

Why

Long-form video is hard to skim, quote, search, or reread. This project turns videos into structured articles you can actually read.

• Local transcription via Whisper — no API cost for audio
• LLM-organized into thematic sections — not a chronological wall of text
• Bilingual out of the box — PT-BR and English with hreflang alternates
• Agent-friendly — Cloudflare Worker on free tier serves Markdown to AI crawlers via content negotiation (75% fewer tokens than HTML)
• Zero frontend build — static HTML deployable to GitHub Pages

Live at [adhenawer.net](https://adhenawer.net/) · [Blog](https://adhenawer.net/blog/)

---

How it works

The pipeline auto-detects the provider from the URL and uses the right strategy to fetch the transcript:

Video URL (YouTube, Twitter/X)
    ↓
src/providers/               — detect provider, capture transcript
  ├── youtube.py                 — captions via youtube-transcript-api
  └──