Task 055: Transcription Data on GitHub Pages

Status: DONE

Motivation

Human directive: 字幕抽出・文字起こしデータは Pages で見られるようにしたい。

Scope

1. Add transcription browsing pages to the site
2. Per-episode transcription page showing:

- Whisper output with timestamps + confidence

- YouTube VTT comparison (where available)

- Speaker attribution status

- Source metadata (model, language, thresholds)

1. Add navigation links from episode pages
2. Include model/condition metadata display

Implementation (2026-02-24)

• TranscriptionPageData type in report-types.ts (assembles lines/dialogue/speakers)
• discoverTranscriptions() in build.ts reads ep*_lines.json + ep*_dialogue.json + ep*_speakers.json
• renderTranscriptionPage() and renderTranscriptionIndex() in templates.ts
• Phase 1 (lines only): timestamp + text + merge reason table
• Phase 2 (attributed): dialogue grouped by scene, speaker names, confidence badges
• Navigation: "文字起こし" link in nav bar + section on index page
• 18 new tests (946 total TS tests, 0 failures)
• All 5 episodes with transcription pages generated

Depends on

• Task 036 (Whisper infrastructure) — DONE
• Task 046 (Whisper EP01-04) — DONE