Redditor drafts yt-dlp + Whisper + SQLite pipeline to auto-summarize expert content
QuestionAsker2030 · reddit · 2026-10-05
A Reddit user wants to use AI to automatically digest multi-source content from 5-10 experts they follow — hour-long YouTube videos, 1,000-word emails, X posts, PDFs — into a private, queryable knowledge base, so they can ask questions like "what's new this month?" or "what are experts recommending?" With a homelab, NAS and basic Linux/Python skills, and while learning Codex/Claude Code, they share a draft pipeline:
- yt-dlp for YouTube metadata → Whisper or YouTube transcripts → cleaned transcript → summary.md → insights. → SQLite + FTS5 full-text index → topic synthesis supporting search, questions and actions.
They ask the community whether this is the best approach and are open to alternatives.
More from coding & agent
- rauchg: handwrite your READMEs for humans, let AI write docs for agents — nicolascraske · 2026-10-05
- Voice-controlled Blender via AI agent: any shortcut triggered by plain English — ctjlewis · 2026-10-05
- 32,000 shards, one loop: procedural Three.js metamorphosis built with Claude Opus — techartist_ · 2026-10-05
- Cloudflare Agents SDK adds Pi Durable harness for crash-resilient long-running agents — mitsuhiko · 2026-10-05
- Codex team pledges daily improvements for 28 days or a full reset — lxfater · 2026-10-05
- OpenAI Agents API in 10 steps: GPT-6.1 Sol nears GPT-6 Astra at one-fifth the price — Roger_M_Taylor · 2026-10-05