Task 004: YouTube Subtitle Collection Script

Status: DONE

Goal

TypeScript script to collect YouTube subtitle/caption data for SOLAR LINE episodes. Define input/output JSON schema (timestamps, speaker, language, source URL).

Details

• Use yt-dlp or YouTube API for subtitle extraction
• Raw data gitignored; only processed/attributed data committed
• Schema defined first (TDD approach)

Depends on

• Task 001 (TS project setup)

Completed Work

• subtitle-types.ts: Two-tier data model (Codex-reviewed)

- Raw: RawSubtitleEntry, RawSubtitleFile (gitignored, fetched via yt-dlp)

- Attributed: DialogueLine, SceneBreak, EpisodeDialogue (committed, human/AI reviewed)

- Speaker registry with canonical IDs and aliases

- OrbitalMention annotations for linking dialogue to TransferAnalysis

- Provenance: schema version, content hash, review metadata

- Integer milliseconds for all timestamps (Codex recommendation)

• subtitle.ts: VTT/SRT parsers, validators, builder utilities
• collect-subtitles.ts: CLI wrapper for yt-dlp (requires yt-dlp in PATH)
• subtitle.test.ts: 25 tests covering parsing, validation, building
• All 137 tests passing (92 TS + 45 Rust), typecheck clean