Video to SOP Generator

Problem It Solves

Teams often have screen-recorded training videos that are hard to turn into searchable, distributable, maintainable documentation. This skill converts a screen recording into a structured standard operating procedure, with HTML, Markdown, and PDF outputs for product launches, onboarding, support, and operations handover.

How It Works

The workflow has four stages. First, scripts/extract_frames.py extracts key frames using --interval for minimum sampling cadence and --threshold for scene-change sensitivity, producing frames_output/ and keyframes_meta.json. Second, scripts/ocr_frames.py recognizes on-screen text with RapidOCR, emitting per-frame text lines and confidence values used to recover menus, buttons, fields, and dropdown labels. Third, the SOP body is drafted against references/sop_spec.md, covering version records, system overview, prerequisites, operation steps, field rules, exception handling, and related operations. Each step should include the user action and the system feedback. Finally, the skill generates multiple formats: HTML with a sticky sidebar table of contents and section anchors, Markdown with standard TOC, tables, and blockquotes, and PDF via scripts/md_to_pdf.py, including CJK font detection, headers and footers, margins, and repeated table headers across pages.

Boundaries and Notes

It works best for recordings with clear UI text and an explicit operation path. Quality depends on OCR for videos that rely on voice narration, complex animation, or have little on-screen text. On Windows, pin onnxruntime==1.20.0; PDF generation requires a CJK font on the system. The three output formats are independent, so editing one format does not automatically update the others.