AI Workflow · Motion
AI User Guide Pipeline
Company
Avada
Timeframe
2026
Role
Creative Design
An AI-assisted production system that turns written help content into scripted, recorded, narrated, and consistently edited videos — with human review before publishing.
The Pain Point
Text documentation is fast to publish but easy to skim and quick to become outdated. Video is clearer, yet the traditional workflow — script, record, trim, zoom, caption, voice, export — makes broad product coverage expensive to maintain.
The Solution
An agent converts each guide into a structured action table; browser automation executes the flow against the real product; Remotion applies reusable zoom, trim, caption, and brand templates; voice generation follows the approved script. Automation handles repetition, while a reviewer remains responsible for the route, sensitive data, narration, and final UI state.
How It Works
Four stages, orchestrated end-to-end: • Claude + Playwright MCP write the interaction script and drive the real app in a browser, recording the screen automatically. • Pipeline + Remotion edit the recording — smooth zoom in / zoom out on key moments, with automatic trimming of loading and idle time. • Auto-diff compares the user-guide docs against the live frontend, flags mismatches, and updates the docs based on what the app actually shows. • ElevenLabs API generates voiceover in sync with the script.
A repeatable output system
The same visual grammar works across different features: consistent framing, purposeful zooms, concise narration, and a branded finish. These sample outputs show how the template adapts while the viewing experience stays familiar.
Result
In the pilot, the workflow produced 20 step-by-step videos in two hours, with one human quality-control pass before publishing. The comparison step also surfaced places where the written guide and live interface had drifted apart, improving documentation accuracy alongside production speed.
