Same Acme contact flow as the first demo,
but the output is a single narrated video instead of an interactive page.
Per-step CSS spotlights highlight whatever the narrator is talking about;
AWS Polly (Arthur, en-GB, neural) reads each step; ffmpeg
composites the title + caption bars in post and stitches the clips.
How it is made
Capture. Playwright drives the app, applies spotlight CSS and screenshots
clean 1920×1080 frames.
Narrate. AWS Polly synthesises one MP3 per step using the Arthur (en-GB)
neural voice.
Overlay. A headless Playwright pass renders the title and caption bar PNGs
from a tiny HTML template.
Stitch. ffmpeg composites each still, both overlays and the matching MP3,
then concatenates the clips.