Markdown

The Markdown export is a .md file: an optional YAML frontmatter block with the video's metadata, an H1 title, then the transcript as paragraphs — or, with timestamps on, sections headed by a clickable timestamp that deep-links back to the video. It drops straight into Obsidian, Notion or Logseq and stays linked to its source.

Frontmatter keys

Scroll the table horizontally →

Key	Value
`title`	Video title (quoted)
`url`	`https://www.youtube.com/watch?v=<id>`
`channel`	Channel name
`published`	Upload date
`duration`	Seconds (number)
`language`	Detected language code
`transcript_source`	"AI Transcription (AssemblyAI)" or "YouTube captions"
`created`	Export date (YYYY-MM-DD)
`type`	Always `youtube`
`tags`	`[youtube, transcript]`

Fields with no value (e.g. a missing channel) are omitted rather than left empty.

Output

With timestamps on, a new ## [HH:MM:SS](youtu.be/<id>?t=N) heading starts wherever there is a gap of more than 5 seconds between segments.

---
title: "Vector Databases Explained"
url: "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
channel: "AI Foundations"
published: "2026-06-01"
duration: 31
language: "en"
transcript_source: "AI Transcription (AssemblyAI)"
created: "2026-07-22"
type: youtube
tags: [youtube, transcript]
---

# Vector Databases Explained

## [00:00:00](https://youtu.be/dQw4w9WgXcQ?t=0)
Welcome to this short introduction to vector databases. A vector database stores embeddings and retrieves them by similarity. …

Sources

INDXR (own code) — frontmatter keys, timestamp-heading format, paragraph splitVerified against packages/shared/src/utils/formatTranscript.ts (generateMarkdown, buildYamlFrontmatter)