How Logseq stores transcripts as Markdown
Logseq watches its pages/ directory for .md files. Drop a converted transcript there (named 2026-02-14 Pricing Discussion.md or whatever convention you use), and Logseq parses every block into a discrete UUID-tagged unit. Speaker headings become block-level headings; paragraphs under each speaker become child blocks. From that point you can [[wikilink]] to the page, ((block-ref)) to a specific quote, or query across with Datalog.
The "voice memo to journal" workflow
Record a voice memo, convert on Audio to Markdown, save into pages/. From your Logseq journal entry for the same day, embed the page ({{embed [[2026-02-14 Pricing Discussion]]}}) or block-reference specific moments (((block-uuid))). The journal stays focused on your thoughts; the source transcript lives in its own page; references connect them with full context.
Splitting long meetings for Logseq performance
Logseq performs best with smaller pages — multi-thousand-block transcripts make the outliner sluggish. For meetings over an hour, split by major topic (use the structured Markdown's ### topic headings as natural cut points), one page per topic, all linked from a parent index page. The graph view then shows the meeting as a small cluster of connected pages instead of one giant node. Pair with PDFs (PDF for Logseq) and web sources (URL for Logseq) for a full-spectrum knowledge graph.