What problem does it solve?
Producing narrated, localized product tutorial videos normally requires manual screen recording, voiceover work, and video editing for every locale and device form factor. This Skill automates the entire pipeline: it drives the real Lotti Linux app under Xvfb with a virtual microphone and live transcription, records the screen, narrates steps with Gemini TTS, fast-forwards waits, composes MP4s via OpenMontage, and uploads results to Cloudflare R2 for docs-site embedding.
Core Features & Use Cases
- Automated video builds: Run make targets to generate tutorial MP4s per locale and per device form factor (desktop or mobile), with TTS narration, cursor glide, HUD clock, and fast-forwarded waits.
- Scenario and locale authoring: Add new tutorial scenarios as YAML configs plus Flutter integration tests, and add locales with per-language narration, voice styles, and dictation text.
- Debugging and publishing: Diagnose failing runs via failure screenshots and diagnostics output, then publish finished videos to Cloudflare R2 with the correct r2.dev URLs for the docs-site TutorialVideo component.
- Use Case: A maintainer needs a German mobile version of the create-task-from-audio tutorial. They run the build with TUTORIAL_LOCALE=de and TUTORIAL_DEVICE=mobile, verify frames with ffmpeg, then publish to R2 with the matching device flag.
Quick Start
Ask the agent to build the tutorial video for a specific scenario and locale, for example: build the create_task_from_audio tutorial video in German for desktop.