Recording

Pensum records audio with the device microphone, creates a meeting note alongside the audio file, and (on stop) hands the result off to the rest of the meeting pipeline: transcription, summarization, action item extraction.

The basic flow

Press Cmd+Shift+R (or run Pensum: Start meeting recording from the command palette). Pensum:

1. Asks for microphone permission the first time
2. Creates a new meeting note in your Meetings folder (Meetings/{year}/{date} {time} Meeting.md by default)
3. Opens the note in the active pane so you can take notes during the call
4. Starts capturing audio in the background

Press Cmd+Shift+R again to stop. Pensum saves the audio as .webm in your Recordings folder, updates the meeting note with the end time, and flips status: in-progress → completed. From there you can transcribe, summarize, and extract action items.

What the note looks like

---
type: meeting
id: 01ABC...
date: 2026-05-16
start_time: "1400"
attendees: []
project:
status: in-progress
recording: Recordings/2026-05-16-1400.webm
ai_summary_status: none
tags: []
---

# Meeting 2026-05-16 1400

## Notes



## Action Items

Fill in attendees and project during or after the call. The downstream features use them:

• Attendees drives speaker identification (mapping Speaker 0/1/2 to actual names) and gives summarization context
• Project lets the meeting roll up into the project dashboard’s open-tasks list

”I’m already in a meeting note — what now?”

Cmd+Shift+R always creates a new note. It doesn’t record into the file you’re currently viewing. There’s no setting in v1.0 that changes this.

A few patterns that work given that constraint:

Situation	What to do
You wrote prep notes in Meetings/Prep — Q3 review.md before the call	When you press Cmd+Shift+R, Pensum creates a fresh Meetings/2026/2026-05-16 1400 Meeting.md. Copy your prep notes into the new note (or link to the prep note via prep: [[Prep — Q3 review]] in frontmatter) and continue from there. The new note is the canonical record because the audio + transcript are bound to it.
You want recurring meetings to land in a specific folder structure	Customize the meeting template at .pensum/templates/meeting.md — change output_folder to e.g. Meetings/1-1s/{{year}}, or rename the file pattern. See the templates guide.
You’re recording a meeting that’s part of a project	Set project: [[Project Name]] in the new note as soon as it opens. The project dashboard picks it up immediately.
You forgot to start recording and the meeting’s already underway	Press Cmd+Shift+R now — you’ll get a partial recording from the moment you started. After stop, you can rename the note to reflect when the meeting actually began.
You want to add audio to an existing meeting note (e.g. you took notes in one note and recorded the audio with a different tool)	Drop the audio file into your Recordings folder, then add recording: Recordings/your-file.webm to the existing note’s frontmatter. Transcription will pick it up from there.

The “always-creates-new” behavior is intentional for v1.0 — it keeps the audio/note pairing unambiguous. A “record into current note” mode is on the post-launch roadmap.

During the recording

The recording runs in the background. You can:

• Take notes in the meeting file as you would normally
• Switch to other notes — recording continues
• Add attendees / project to the frontmatter
• Switch apps — the browser keeps the recording active as long as Obsidian stays open

If you close Obsidian mid-recording, the audio buffer is lost. Pensum saves chunks every 5 seconds internally, but no partial-recovery is wired up in v1.0. Don’t close Obsidian until you’ve stopped the recording.

After you stop

Pensum:

1. Writes the audio to your Recordings folder
2. Updates the meeting note’s status to completed and adds end_time
3. Shows a notice with the duration

The note is now ready for the rest of the pipeline. Open the Meeting Page sidebar view (or the buttons inside the note) for:

• Transcribe — sends the audio to Deepgram (managed) or your chosen provider (BYO), saves a .transcript.json file alongside the audio. See transcription.
• Summarize — generates a summary callout inside the note from the transcript. See summarization.
• Extract Action Items — pulls action items out of the transcript and appends them as task checkboxes. See action items.
• Identify Speakers — maps Speaker 0/1/2 labels to actual names using your attendees list.

You can run these in any order. Most users transcribe first, then summarize and extract.

Where things live

Setting	Default	What it controls
Meetings folder	Meetings	Meeting notes; year subfolder added automatically (Meetings/2026/...)
Recordings folder	Recordings	Audio files
Audio retention	Keep indefinitely	Optional auto-delete after 30 or 90 days (the note + transcript stay)

All configurable under Settings → General (folders) and Settings → Meetings (retention).

Mobile

On Obsidian Mobile (iOS / Android), recording uses the WebView’s MediaRecorder API. Audio quality depends on the device. The mobile flow is identical to desktop: Cmd/Tap the command, take notes, stop.

A few things to know on mobile:

• Transcription is deferred to desktop by default. The plugin captures the audio file to your vault; the next time you open the vault on a desktop with Pro, you can run Transcribe on the meeting note.
• Background recording is constrained. If you switch apps or lock the device, iOS / Android may suspend the recording. The audio that was captured up to that point is still saved.
• Native widgets, share extensions, watch surfaces aren’t available — Obsidian Mobile plugins can’t expose those. If you need rich native recording surfaces, that’s part of the future native standalone app, not v1.0.

Privacy

• Audio stays in your vault. Pensum does not upload audio anywhere unless you explicitly trigger transcription.
• On the Pro All-in-One plan, transcription audio passes through Pensum’s worker transiently and is not retained. On BYO, audio goes directly to your chosen provider (Deepgram, OpenAI). See transcription and the architecture page for the full data flow.