Meetings Ai for Obsidian

Obsidian plugin that records, transcribes, and summarizes meetings directly inside your vault. One click to start recording, one click to get a structured summary with action items.

How it works

1. Click the microphone icon or run the command to create a meeting note
2. Press Start to record
3. Press Stop when done — the plugin sends audio to OpenAI Whisper, gets a transcript with timestamps, then generates a structured summary

Everything stays in your vault: the audio file, transcript, and summary are saved as part of the note.

Features

Interactive widget — a self-contained meeting panel embedded in your note with tabs for Summary, Notes, Transcript, and Audio.

Structured summaries — the AI outputs key points, decisions, and action items. Action items render as checkboxes you can tick off.

Footnote references — each summary point has clickable badges like {1,2,3} that jump to the exact transcript segments it was based on.

Timestamped transcript — Whisper returns segments with start/end times so you can see when each part of the conversation happened.

Error recovery — if the network drops, the audio is already saved locally. A retry button lets you re-send for transcription when you're back online. This state persists even if you switch pages or restart Obsidian.

Meeting description — add a short description to each meeting for context.

Auto-archive — meetings from previous months are automatically moved into YYYY-MM subfolders.

Works on iPad/iOS — recording uses audio/mp4 fallback for Safari compatibility.

Setup

1. Install the plugin
2. Go to Settings > Meetings Ai
3. Add your OpenAI API key
4. (Optional) Configure output folder, audio folder, summary prompt, and assistants

Settings

Setting	Description	Default
OpenAI API Key	Required for transcription and summarization	—
Save audio	Keep audio files in your vault	true
Link audio	Embed audio link in the note	true
Output folder	Where meeting notes are created	Obsidian default
Audio folder	Where audio files are saved	Same as output
Date format	Format for note titles	DD.MM.YY
Transcription hint	Words/names to help Whisper spell correctly	—
Summary model	OpenAI model for summarization	gpt-4o
Assistants	Custom prompt profiles for different meeting types	1 default

Credits

This project is a fork of Magic Mic by Drew McDonald, which provided the original recording and transcription foundation.

License

MIT