Audio to Markdown for Obsidian: Voice Notes to Your Vault
Obsidian was built for written notes. Most thinking happens in voice: meetings, walking notes, dictated drafts, podcast reactions you don't want to lose. Transcribe the audio, convert to structured Markdown with topic sections and timestamps, drop into the vault. Voice notes become first-class citizens of the graph: searchable, linkable, taggable, surfaced by backlinks.
What changes when voice memos become Markdown notes
An audio file in your vault is a dead-end attachment: Obsidian stores it but doesn't index the words, can't [[wikilink]] to it, and the graph view shows it as an orphan. Convert to Markdown and the same content becomes a real note: full-text search, atomic-note splitting, YAML frontmatter for metadata, tags, and graph view connections to anything you reference.
The "voice memo to atomic note" workflow
Record a voice memo, convert to Markdown via Audio to Markdown, save into the vault with date-prefixed naming (2026-02-14 Voice Memo - Pricing Model.md) and YAML frontmatter capturing recording date, duration, and any tags you want indexed. The body is the structured transcript: continuous prose, punctuated and paragraphed, with H2 headings wherever the topic shifts.
Then split the long memo into atomic notes manually or with the Note Refactor plugin: one idea per note, each backlinking to the source memo. Your Zettelkasten now ingests spoken thought the same way it ingests written thought. Also clip web pages (URL for Obsidian) and import PDFs (PDF for Obsidian) for a full-spectrum vault.
Frontmatter template for voice content
Common fields: type: voice-memo, recorded: 2026-02-14, duration: 14:22, tags: [pricing, q2-planning], participants: [Sarah, Marcus], source-audio: ./attachments/2026-02-14-pricing.m4a. Dataview can then query across all voice memos by participant, date, or tag.
Frequently asked questions
Why convert voice memos to Markdown instead of keeping audio in the vault?
Obsidian doesn't index audio file content: no full-text search, no [[wikilinks]] to specific moments, no graph-view connections. A Markdown transcript is a real note: searchable, linkable, splittable into atomic notes. Keep the original audio as an attachment too if you want, and reference it from the transcript's frontmatter.
How do I add YAML frontmatter to a converted voice memo?
Top of the .md file, between two --- lines. Common keys for voice content: type, recorded (date), duration, tags, participants, source-audio (path to the original file). Obsidian indexes everything in frontmatter for queries, Dataview, and the file-properties panel.
Can I split a long meeting transcript into atomic Zettelkasten notes?
Yes: that's often the goal. Convert first to get the structured Markdown, then either split manually by topic or use Note Refactor to break the long note into linked sub-notes preserving wikilink integrity. Each atomic note represents one idea from the meeting; backlinks point to the source transcript.
Does the graph view show voice memos and their connections?
Once converted to .md, yes. Each transcript is a node. Atomic notes derived from the memo are linked nodes. Tags cluster nodes visually. Wikilinks from your permanent notes back to memo passages create the same kind of dense graph you get from text-only Zettelkastens, but now with spoken thought feeding the system.
Best plugins for voice memos in Obsidian?
Dataview (querying memo metadata), Note Refactor (splitting long transcripts into atomic notes), Templater (auto-generating frontmatter on import), Smart Connections (semantic search across the vault including transcripts), Audio Recorder if you want to record directly into the vault. Combined with structured Markdown conversion, these turn Obsidian into a credible voice-PKM system.