Obsidian Meeting Notes on Mac: Transcribe Calls Into Your Vault

If your knowledge base lives in Obsidian, your meeting notes are probably the leakiest part of it. The meetings happen in Zoom, Teams, or Meet; the notes — if they exist at all — end up in whatever tool captured them: an AI notetaker’s web dashboard, a Google Doc someone shared, or a hurried bullet list you typed while half-listening. None of it lands in the vault, so none of it links, none of it surfaces in search, and six months later the decision you need to reconstruct lives in some notetaker’s cloud account instead of next to the project note where it belongs.

The fix is a pipeline: record the meeting on your Mac, transcribe and summarize it locally, and export Obsidian meeting notes as Markdown that drop straight into the vault. As of macOS 26 the entire pipeline runs on-device — which matters double for Obsidian users, because the reason you chose Obsidian is the same reason you shouldn’t route your meetings through someone else’s server.

Files over app, audio included

Obsidian’s core bargain is files over app: your notes are plain Markdown files in a folder you own. No proprietary database, no vendor cloud, no export lock-in. People pick it exactly because the knowledge outlives the tool.

Cloud meeting notetakers are the opposite bargain. Otter, Fireflies, Fathom, and the rest capture your meetings into their systems — audio on their servers, transcripts behind their login, summaries shaped by their retention and training policies. Even when export exists, the workflow drags you through a web app to get your own words back. Bolting a cloud-silo notetaker onto a local-first vault is philosophically incoherent, and practically it means your meeting knowledge lives in the one place your linked, searchable, backed-up system can’t reach.

A local pipeline keeps the bargain intact end to end: the audio is recorded on your Mac, the transcription runs on your Mac, and the artifact is a Markdown file in a folder — exactly the shape Obsidian wants.

The pipeline: call → transcript → vault

Here’s the full workflow with Dictanta, which handles the recording, transcription, and summarization on-device:

Record the meeting with no bot in the call. Dictanta captures your Mac’s system audio — the far end of the Zoom, Teams, Meet, or Webex call — plus your microphone. There’s no “Notetaker has joined the meeting” participant, no admission prompt for the host, nothing your counterpart sees. How this works under the hood is covered in recording system audio on a Mac.
Transcription runs locally when the call ends. Apple’s SpeechAnalyzer framework converts the recording to text on the Neural Engine — no upload, no processing queue, roughly 55% faster than Whisper v3 Turbo on the same hardware per Apple’s benchmarks.
The summary is generated on-device too. Apple’s Foundation Models produce a short summary, key points, and action items from the transcript. Every bullet is audio-anchored: inside Dictanta, tapping a summary point jumps playback to the exact moment it was said, so you can verify a claim against the actual words in seconds.
Export Markdown into the vault. The export is a single .md file — title, summary, key points, action items, full transcript underneath. Save it directly into a folder inside your vault. Because an Obsidian vault is just a directory, the note appears in Obsidian instantly; there is no import step, no plugin, no sync bridge.

That last point deserves emphasis because it’s the reason this workflow is so short: getting a file into Obsidian is saving a file into a folder. The same property that makes Obsidian local-first makes it trivially interoperable with any tool that writes Markdown.

What the note looks like

A 30-minute project call exports as something like this:

# Q3 roadmap sync — 2026-07-13

## Summary
Reviewed the Q3 launch scope with the design and API teams.
Design is signed off; the API migration slips one sprint.
Agreed to cut the analytics rework to protect the date.

## Key points
- Design sign-off complete, no open items
- API migration is the critical path, one sprint behind
- Analytics rework moves to v1.1

## Action items
- Me: confirm revised API timeline by Friday
- Dana: update the launch checklist

## Transcript
Okay, so the main thing on the launch is the API migration,
everything else is basically ready to go...

From there it’s a normal Obsidian note, and a few minutes of light editing wires it into the graph: wrap names and projects in [[wikilinks]] ([[Dana]], [[Q3 launch]]), add tags or properties to match your setup, and pull the action items into wherever your tasks live — checkboxes in a daily note, a Tasks-plugin query, whatever you already use. The heavy lifting — turning 30 minutes of speech into structured text — is already done; you’re just linking.

Vault patterns that work for meeting notes

Three arrangements, in increasing order of ceremony:

The inbox folder. Export every meeting into Meetings/Inbox/, and process notes into their project homes during a weekly review. Lowest friction; the transcript is captured and searchable immediately even if you never file it.

One note per meeting, linked from the project. File each export as Meetings/2026-07-13 Q3 roadmap sync.md and add a link from the relevant project note. Six months later, the project note reads as a chronology: every decision, linked to the meeting where it happened, with the verbatim transcript one click deeper.

Daily-note embedding. If your vault revolves around daily notes, link each meeting from the day it happened. Combined with backlinks from people and project pages, “what did we discuss with the vendor in May” becomes a ten-second lookup instead of an archaeology dig.

Whichever pattern you use, full-text search over transcripts is the quiet superpower. The exact phrase someone used, the number they quoted, the caveat they slipped in — all of it is grep-able plain text in your vault, forever, at no per-seat price.

Why not just use a transcription plugin inside Obsidian?

Community plugins exist that transcribe audio files from within Obsidian, and for short voice notes they can be fine. For meetings they fall short in two structural ways. First, they start from an audio file — they don’t solve capture. Recording both sides of a Zoom call on a Mac requires system-audio capture, which is an app-level capability, not something a text editor’s plugin does; without it there’s no meeting audio to transcribe in the first place. Second, the popular plugins mostly delegate to a cloud transcription API, which reintroduces exactly the audio-leaves-your-machine problem the local-first stack was supposed to avoid — and self-hosted alternatives trade that for setup and maintenance work.

A dedicated on-device app in front of the vault solves capture, transcription, and summarization in one step, and hands Obsidian what Obsidian is best at holding: the Markdown.

Why not Granola or Otter with an export step?

Granola is the closest cloud comparison, since it also skips the bot and produces polished notes. But the processing happens on Granola’s servers, the notes live in Granola’s system, and getting them into Obsidian is a recurring copy-paste chore against the grain of the product. Otter is bot-based, cloud-based, and dashboard-centric — three steps removed from a Markdown file in a folder. The detailed comparisons are in the Granola alternative and Otter alternative posts; the short version for Obsidian users is that both put a cloud silo between your meetings and your vault, and the export step never stops being manual.

There’s also the participant-facing cost: a bot in the attendee list changes the meeting’s temperature, and some hosts simply eject them. No-bot capture on your own Mac sidesteps the whole category of problem — covered in depth in the meeting recorder with no bot guide.

The privacy line, drawn concretely

It’s worth being precise about what “on-device” buys you here, because Obsidian users tend to care about the specifics:

The audio never leaves the Mac. Capture, transcription, and summarization all run locally on Apple silicon. There is no server-side processing step anywhere in the pipeline.
No account holds your content. There’s no web dashboard where your meetings accumulate. The recording lives in the app; the exported note lives in your vault.
The Markdown is yours in the strongest sense. It syncs however your vault syncs — iCloud, Obsidian Sync (end-to-end encrypted), Syncthing, git — under rules you chose.

For confidential calls — client work, personnel discussions, anything under NDA — this is the difference between “we should check the vendor’s data processing terms” and “there is no vendor.” The full argument is laid out in private transcription on the Mac.

Voice notes feed the same vault

Meetings are the highest-volume source, but the same pipeline handles the rest of your audio inbox: a Voice Memo dictated on the walk home imports and exports through the identical on-device flow, landing in the vault next to the meeting notes. That workflow is covered in transcribing Voice Memos to Markdown. One pipeline, one output format, one vault.

Bottom line

If Obsidian is your source of truth, meetings shouldn’t be the exception. The workflow that fixes it is short: Dictanta records the call on your Mac with no bot, transcribes and summarizes it on-device, and exports a structured Markdown note you save into the vault — where it links, searches, and outlives every tool involved. No cloud silo, no copy-paste ritual, no audio on anyone’s server.

Dictanta is free for your first three recordings with no length cap — enough to run a few real meetings through the pipeline and see the Markdown land in your vault before paying anything. Paid tiers are $9.99/mo, $79.99/yr, or $149.99 lifetime.

FAQ

How do I get a meeting transcript into Obsidian?

Export the transcript as a Markdown file and save it into any folder inside your vault. Because an Obsidian vault is just a folder of files, the note appears immediately — no import step or plugin needed. Dictanta exports meetings as structured .md (summary, key points, action items, transcript) for exactly this.

Can I transcribe meetings for Obsidian without sending audio to a cloud?

Yes. On macOS 26, Apple’s SpeechAnalyzer framework and Foundation Models run transcription and summarization entirely on-device. Dictanta uses both, so the audio and transcript never leave your Mac — the only output is the Markdown file you save into your vault.

Does this work for Zoom, Teams, and Google Meet?

Yes. The capture is platform-agnostic: Dictanta records the Mac’s system audio plus your microphone, so any call that plays through your Mac — Zoom, Teams, Meet, Webex, or a plain phone call routed through the Mac — is recorded the same way, with no bot joining the meeting.