Voice Notes (Local Whisper)

Record voice notes and meetings with local Whisper transcription, speaker identification, and AI-powered meeting summaries.

Features

• Dictation mode - real-time voice-to-text at your cursor position
• Full voice notes - record, transcribe, and save with audio playback
• Meeting mode - capture system audio + mic with live transcript sidebar
• Speaker labels - instant "Me" vs "Speaker" identification (macOS with ScreenCaptureKit)
• AI meeting notes - summaries, action items, and key decisions
• Fully local - no cloud transcription service, your audio never leaves your machine

Requirements

• A Whisper transcription server running locally or via Docker
• macOS (Apple Silicon recommended) or any OS with Docker
• See voice-notes-server for server setup

Quick start

1. Install this plugin from Obsidian Community Plugins
2. Set up the transcription server (instructions)
3. In plugin settings, verify Server URL is http://127.0.0.1:5678
4. Click the microphone ribbon icon or run Start dictation

Commands

Command	Description
Toggle dictation	Start or stop real-time voice-to-text
Record full voice note	Open modal for record/stop/save workflow
Start meeting transcription	Begin meeting mode with system audio capture
Stop meeting transcription	End meeting and generate notes
Mark moment in meeting	Bookmark a moment during a meeting
Generate meeting notes from selection	Summarize selected text with AI
Check Whisper server status	Verify server connectivity

Meeting mode

Meeting mode captures both system audio (the other speakers) and your microphone, transcribes in real-time to a sidebar panel, and generates structured notes when you stop.

Speaker identification: On macOS 14+ with ScreenCaptureKit, speakers are labeled instantly by comparing system audio energy (them) vs mic energy (you). No server processing required.

Post-meeting options (configurable in settings):

• Transcript only
• AI summary with key decisions
• Full notes with action items, decisions, and follow-ups

AI summarization

Supports three providers for meeting summaries and action extraction:

• Anthropic (Claude)
• OpenAI (GPT)
• Ollama (local, free)

Configure your provider and API key in Settings.

Server setup

This plugin requires a companion transcription server. See voice-notes-server for:

• Native macOS (Apple Silicon MLX GPU) - fastest option
• Docker (any OS, CPU) - portable option