EchoNotes

Private meeting transcription for Mac. No bots. No cloud. No subscriptions.

EchoNotes is a native macOS desktop app that records audio from any call — Zoom, Google Meet, FaceTime, Discord, phone calls, whatever — and transcribes it locally using WhisperKit. Optionally generate AI-powered meeting summaries with your existing ChatGPT subscription or any API key.

🎤 AUDIO-ONLY APP — We capture system audio output (other people on calls) and microphone input (your voice). No video, screen recording, or visual data whatsoever — and no Screen Recording permission either: system audio comes from a Core Audio process tap.

Features

• Universal recording — works with any app that plays audio through your system
• Live transcription — see text appear in ~5-second intervals while you record
• Post-recording transcription — transcribe the full audio after you stop for higher accuracy
• Speaker diarization — distinguish between different speakers in the transcript
• Recording library — browse, search, and manage all your recordings with full-text search
• AI summaries — generate structured meeting summaries (key points, action items, decisions)
• Multi-provider AI — OpenAI, Anthropic, Google Gemini, or local Ollama
• ChatGPT OAuth — sign in with your existing ChatGPT Plus/Pro subscription (no API key needed)
• Global hotkey — start/stop recording with ⌘⇧R from anywhere
• 100% local transcription — all speech-to-text happens on-device via Apple's Neural Engine
• Free — no accounts or usage limits for recording and transcription

How It Works

Core Audio process tap (system audio) ──→ system.caf ──┐
                                                        ├──→ session folder + meta.json
AVAudioEngine (microphone)            ──→ mic.caf    ──┘
                                                            │
                                              WhisperKit (CoreML, per track)
                                                            │
                                              merged Transcript ("You"/"Them")
                                              (transcript.txt + transcript.json)
                                                            │
                                                  AI Provider (optional)
                                                         │
                                                 Meeting Summary
                                          (key points, actions, decisions)

Requirements

• macOS 15.0+ (Sequoia or later — system audio capture uses Core Audio process taps)
• Apple Silicon (M1 or later recommended for fast transcription)

Build & Run

git clone <repo>
cd echonotes

Build the app (recommended)

./scripts/build-app.sh

This creates EchoNotes.app in the project root. Double-click to launch, or drag to /Applications to install.

Run from Xcode

open Package.swift

Press ⌘R to build and run.

Run from command line (debug)

swift build
.build/debug/EchoNotes

On first launch, the app downloads a Whisper model (~150MB for Base English). This only happens once.

Permissions

macOS will ask for:

• Microphone — to capture your voice
• System Audio Recording Only — to capture system audio output (prompted on first recording)

No Screen Recording permission is needed: system audio is captured through a Core Audio process tap, which has its own audio-only permission under System Settings → Privacy & Security → Screen & System Audio Recording.

Usage

1. Launch EchoNotes — a window opens with a sidebar and detail area
2. Choose transcription mode in the toolbar: Live or After Recording
3. Start a call in any app
4. Click Record in the toolbar (or press ⌘⇧R)
5. Click Stop when done
6. Your recording (mic + system tracks) + transcript are saved to a session folder in ~/Documents/EchoNotes/
7. Select a recording in the sidebar to view the transcript
8. Click Ask EchoNotes to generate an AI summary
9. Open Settings (⌘,) to configure AI provider, Whisper model, and transcription defaults

AI Summarization

EchoNotes can generate structured meeting summaries from your transcripts. Configure in Settings (⌘,) → AI tab:

Option 1: ChatGPT Sign-In (no API key needed)

Click Sign in with ChatGPT to use your existing Plus/Pro subscription. Uses OAuth 2.1 + PKCE to authenticate, then calls the ChatGPT backend Responses API directly.

Option 2: API Key

Paste an API key for any supported provider:

• OpenAI — api.openai.com
• Anthropic — Claude models
• Google Gemini — Gemini models
• Ollama — local models (no key needed)

Project Structure

EchoNotes/
├── AI/
│   ├── AIProvider.swift             # Multi-provider config (OpenAI, Anthropic, Google, Ollama)
│   └── AIService.swift              # Summarization logic + ChatGPT backend streaming
├── App/
│   ├── EchoNotesApp.swift           # Entry point (WindowGroup + Settings scene)
│   ├── AppDelegate.swift            # Owns shared state (RecordingEngine, RecordingLibrary)
│   └── RecordingEngine.swift        # Coordinates capture + writing + transcription
├── Audio/
│   ├── SystemAudioCapture.swift     # ScreenCaptureKit (audio-only)
│   ├── MicrophoneCapture.swift      # AVAudioEngine mic input
│   └── AudioFileWriter.swift        # Stereo M4A output
├── Auth/
│   ├── OAuthManager.swift           # OAuth 2.1 + PKCE flow, JWT parsing, token storage
│   └── OAuthCallbackServer.swift    # Local HTTP server for OAuth redirect
├── Models/
│   ├── Recording.swift              # Recording model + library management
│   ├── Speaker.swift                # Speaker enum for diarization
│   └── Transcript.swift             # Segment model + export (.txt, .json)
├── Transcription/
│   ├── WhisperEngine.swift          # WhisperKit wrapper
│   ├── StreamingTranscriber.swift   # Live transcription (5s chunks)
│   ├── TranscriptionManager.swift   # Orchestrates transcription + AI config + OAuth state
│   └── ModelManager.swift           # Model loading + caching
├── Views/
│   ├── MainWindowView.swift         # Root NavigationSplitView (sidebar + detail)
│   ├── SidebarView.swift            # Sidebar meeting list with search
│   ├── ActiveRecordingView.swift    # Recording session UI (timer, levels, live transcript)
│   ├── RecordingDetailView.swift    # Transcript + AI summary with DisclosureGroups
│   ├── DesktopSettingsView.swift    # Tabbed Settings window (General, AI, About)
│   ├── SettingsView.swift           # AI provider config + OAuth sign-in
│   ├── LibraryView.swift            # Recording library (date formatter shared)
│   ├── RecordingControlsView.swift  # Start/stop button
│   ├── SummaryView.swift            # Meeting summary display
│   ├── TranscriptDisplayView.swift  # Completed transcript display
│   └── LevelMeterView.swift         # Audio level visualization
├── Utils/
│   ├── Constants.swift              # Audio configuration
│   └── Permissions.swift            # Mic + screen recording checks
├── scripts/
│   └── build-app.sh                 # Build release app bundle
├── research/
│   └── competitive-analysis.md      # Market research
├── CLAUDE.md                        # AI coding assistant rules
├── CONTRIBUTING.md                  # Build + test requirements for PRs
└── EchoNotesTests/                  # ~90 tests across 10 files

Tech Stack

• Swift / SwiftUI — native macOS desktop app with NavigationSplitView
• SwiftPM — package management (no Xcode project file needed)
• ScreenCaptureKit — system audio capture (audio-only mode)
• AVAudioEngine — microphone capture
• WhisperKit — on-device speech-to-text via CoreML + Apple Neural Engine
• Network.framework — OAuth callback server (NWListener)
• AAC/M4A — compressed audio output

Contributing

See CONTRIBUTING.md. Every PR must:

1. Pass ./scripts/build-app.sh
2. Include tests for new functionality
3. Use feature branches (never push directly to main)

License

Proprietary — All rights reserved.