CCTV Tutorial Builder
ToolA pipeline that generates tutorial videos deterministically from a recipe file — capturing real device interfaces over both web and HDMI, validating every asset, and rendering the finished video without a human ever recording a screen.
Overview
Recording product tutorials by hand does not scale. Every firmware revision invalidates the footage, every brand needs its own terminology, and every language means re-recording from scratch. The result is a support library that is permanently out of date.
This pipeline treats a tutorial as data. A recipe file declares the steps, the target surface, the brand, and the device. The CLI then captures the required assets — Playwright drives real web interfaces, while a Python toolchain captures HDMI output for device menus that have no web UI at all — validates that every declared screenshot actually exists, and hands off to a Remotion compositor for rendering.
Brand adapters translate terminology per manufacturer, so one recipe produces correct videos for multiple brands. A crawl mode walks an entire device UI using adapter hints and emits a manifest, which turns authoring new recipes from a discovery problem into an editing one.
Key Features
- Recipes as declarative JSON with a validating schema
- Playwright capture pipeline for web-based device interfaces
- Python HDMI capture for device menus with no web UI
- Remotion compositor for deterministic video rendering
- Brand adapters translating terminology across manufacturers
- Crawl mode that walks an entire device UI and emits a manifest
- Asset presence validation before a render is allowed to start
- Unified CLI orchestrating capture, validate, render, and publish
- Multi-language and multi-target output from a single recipe
Tech Stack
Architecture Highlights
Content that regenerates itself?
If your documentation goes stale every release, I can build a pipeline that regenerates it instead of a team that re-records it.
Start a Conversation