Visit Website

🚀 We're on Product Hunt! If you find Note67 useful, please consider upvoting us on Product Hunt — it helps others discover the project!

A private, local meeting notes assistant. Capture audio, transcribe locally with Whisper, and generate AI-powered summaries — all on your device.

• Meeting detection can be turned off in Settings > System — it prompts once per call, never during a recording, and not just because Teams or Discord is open
• The meeting prompt shows even when Note67 is closed to the menu bar, and goes away a couple of minutes after the call ends
• Uploaded audio is transcribed as soon as it's added, and stopping a recording while an upload is transcribing no longer skips the recording's final transcription
• Deleting a task can be undone for a few seconds, subtasks included, and deleting a note tells you how many tasks go with it
• Overdue tasks are highlighted in a note's Tasks tab too, and task checkboxes work with the keyboard and screen readers
• Links in summaries, Settings and first-run setup open in your browser (the Settings and setup links had done nothing since 0.1.22)
• When Retranscribe fails, Note67 says why and keeps the transcript you had
• Fixed Note67's window not coming back after it started at login, and clicking through a task's details marking it as edited
• Security — Note67's window reaches only your recordings and the files you choose to save, and HTML in a note, transcript or AI summary can't run code or send the window elsewhere
• Privacy — images from the web no longer load in notes or summaries, so an AI summary can't use one to send your notes somewhere; images you paste still show
• Security — updated the secure-connection library the updater uses for a security advisory
• About 2,500 lines of unused code and five unused libraries removed

• Fixed the app quitting after a recording when the summary contained Chinese, Japanese or Korean text, or an emoji or em dash in the wrong place
• Fixed the app quitting when opening a note that another note mentions near non-English text (the "unlinked mentions" panel behind it is removed)
• Fixed microphone-only recordings ending with an empty transcript and losing everything recorded before a pause, and continuing an older recording overwriting it
• Non-English meetings are no longer re-transcribed in English after Stop — your transcription language is used everywhere
• Notes typed during a meeting are saved as you type and before the summary is written, and clearing a note's text now sticks
• Only one recording runs at a time, so a second one can no longer file a meeting's audio under the wrong note
• Recording no longer needs Ollama — a transcription model is enough, and Ollama adds summaries, titles and tasks when it's running
• Summarize no longer renames notes you've titled yourself, and search works with hyphens, apostrophes, dots and email addresses
• Finding tasks reads the whole of a long meeting without adding the same task twice, and tasks you add yourself no longer turn overdue the next day
• Uploaded audio plays and transcribes on notes without a recording, and mono .m4a files such as voice memos convert at full length
• Right-click in the editor shows the normal Cut, Copy, Paste and spelling menu
• Coming in 0.2 — Note67 focuses on meeting notes: the graph view, hashtags, backlinks, the Cmd+J writing sidebar and PDF export are going away (Markdown export stays), and nothing you wrote is deleted

• Security — updated bundled third-party components to patch five known vulnerabilities, including the sanitizer that cleans pasted content in the editor

• Fixed your own voice missing from the transcript — quiet microphones are now boosted before transcription instead of being treated as silence
• Fixed short replies like "hello" or "thanks" being dropped from the live transcript
• Fixed speech being discarded when you talk at the same time as the meeting audio
• Whisper's stage directions (*Slow's voice*, *sighs*) no longer show up as transcript lines
• Live transcription now appears below earlier recordings instead of jumping to the top, and stays in view as it arrives
• Fixed silent playback audio when continuing a recording
• Retranscription no longer loses your existing transcript if it's interrupted
• Recommended transcription model is now Turbo Q8 — same quality and speed, half the download
• Faster startup and a smaller app bundle

• Tasks — turn a meeting's action items into a real to-do list, or add standalone tasks; each task has a due date, description, and subtasks
• Central Tasks page — open tasks from every meeting in one place, grouped by date (Overdue / Today / Tomorrow / Future) with quick filters
• AI "Generate" pulls action items straight from a meeting into your task list; click a task to jump back to its note
• Smarter summaries — when you take your own notes, the AI summary now expands them with detail from the transcript instead of ignoring them
• First-run setup wizard — guided setup for the transcription model, AI assistant, and microphone/screen permissions

• Sentence-based transcript formatting — long monologues break into readable sentences instead of one block of text
• Interval timestamps — a timestamp every 30 seconds within a speaker turn for easier navigation
• Fixed transcript ordering — the last line no longer jumps to the top after post-recording enhancement
• Smarter noise filtering — drops Whisper hallucinations (stray punctuation, repeated "Thank you", stuck words) on silence
• Improved echo handling — better removal of mic pickup of speaker audio when not using headphones
• Transcript search no longer breaks on special characters

• Listen-only recording mode — capture system audio without microphone
• Record meetings when mic is unavailable or only system audio is needed
• Live transcription support in listen-only mode
• Improved loading indicator during post-recording retranscription

• Wiki-style links — type [[Note Title]] to link between notes
• Link aliases — use [[Title|display text]] syntax for custom link text
• Link preview on hover — see snippet, click to navigate
• Link autocomplete — type [[ to see note suggestions
• Backlinks panel — see which notes link to the current note
• Unlinked mentions — find notes that mention this note's title without [[]]
• Auto-update links when note title changes
• Hashtag support — type #tag in notes for auto-extraction and filtering
• Tag autocomplete — type # to see suggestions with keyboard navigation
• Auto-generated tag colors based on tag name
• Global search (Cmd+K) — full-text search across all notes with highlights
• Updated recommended Ollama model to Gemma 4

• Whisper large-v3-turbo model — 8x faster, similar accuracy to large-v3
• Quantized model variants (Q8) — smaller downloads, lower memory usage
• Changed recommended model to large-turbo for better transcription quality

• Minor bug fixes and performance improvements

• AI writing assistant sidebar — improve, summarize, expand, or rewrite notes with AI
• Quick actions: Summarize, Action Items, Improve, Expand, Fix Grammar, Bullets
• Chat interface for custom AI requests
• Insert or replace note content with AI-generated text
• Keyboard shortcut Cmd+J to toggle AI sidebar

• Meeting management (create, end, delete)
• SQLite database for local storage
• Audio recording (microphone)
• Local transcription with Whisper
• Speaker distinction (You vs Others) on macOS
• Echo deduplication for speaker usage
• Live transcription during recording
• Auto-retranscribe after recording for improved accuracy
• Pause/Resume recording
• Continue recording on existing notes (Listen)
• Listen-only mode (system audio without microphone)
• Upload external audio files for transcription (mp3, m4a, aac, flac, ogg, wav, mkv, mp4; Opus audio isn't supported yet)
• Download and delete audio files
• Reorder audio files with up/down controls
• Retranscribe audio with different Whisper model
• Voice Activity Detection (VAD) for mic input
• Automatic filtering of blank/noise segments
• Transcript viewer with search and speaker filter
• AI-powered summaries via Ollama
• Smarter summaries that expand your own notes using the transcript
• AI writing assistant sidebar (improve, summarize, expand notes)
• Tasks with due dates, descriptions, and subtasks
• Central Tasks page grouped by date (Overdue / Today / Tomorrow / Future)
• AI-generated action items from meetings into your task list
• First-run setup wizard (model, AI assistant, permissions)
• Export to Markdown
• Settings with Profile, Whisper, Ollama, System tabs
• Dark mode support
• Custom context menus
• System tray support
• Cross-platform system audio (Windows via WASAPI)
• Rich markdown editor with live preview (Notion-style)
• Local image storage with paste support
• Slash commands for quick formatting
• LaTeX/math support in notes
• Wiki-style links with [[Note Title]] syntax
• Link aliases, preview on hover, and autocomplete
• Backlinks panel
• Hashtag support with autocomplete and auto-generated colors
• Global search (Cmd+K) with full-text search
• Linux system audio support

Light Mode	Dark Mode	Settings

AI Summary

Note67 can distinguish between your voice and other meeting participants:

• macOS 13.0 (Ventura) or later
• Screen Recording permission (System Settings → Privacy & Security → Screen Recording)
• Microphone permission

• Windows 10 or later
• Microphone permission
• No additional permissions needed for system audio (WASAPI loopback)

When using speakers instead of headphones, your microphone picks up audio from your speakers, causing duplicate transcriptions. Note67 handles this with a multi-layer approach:

How it works:

1. Level normalization + Voice Activity Detection (VAD) - Quiet microphones are boosted to a usable level first, then transcribed only if their energy clears a threshold. Normalizing before the check is what keeps a quiet speaker from being mistaken for silence
2. Echo Deduplication - Mic transcripts are compared against a 30-second rolling history of system audio segments
3. Text Similarity Matching - A mic segment is treated as echo only when it repeats words from an overlapping system segment. Overlapping in time is not enough on its own — two people talking at once is normal in a meeting, so that speech is kept

For best results:

• Headphones are still recommended for optimal quality
• Works automatically when system audio capture is enabled

• Node.js (v18+)
• Rust
• Ollama (for AI summaries)

# Install Rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# Install Tauri CLI
cargo install tauri-cli

# Install Ollama and pull a model
brew install ollama
ollama pull llama3.2

# Install dependencies
npm install

# Run dev server (opens app window)
npm run tauri dev

# Build for production
npm run tauri build

Note: Windows system audio capture via WASAPI loopback does not require additional permissions.

AGPL-3.0