This example converts a JSON array of audio annotations—each with start and end times and optional notes—into WebVTT subtitle cues. It defines a `WebVTT` constructor that formats timestamps from seconds into `HH:MM:SS` and builds cue blocks with the `annotation2cue` method. The sample data contains Japanese text notes, including lines from "Rashomon," which become cue content. The conversion runs through `forEach` over the JSON and logs the resulting WebVTT string to the console, demonstrating a lightweight, dependency-free approach to generating subtitle files from structured annotation data.