EchoNote

Private, on-device meeting transcription & AI summaries.
No cloud. No bots. No subscriptions.

Quick Start

1. Download the installer for your OS → Latest Release
2. Install & open the app — a setup wizard walks you through permissions and model downloads.
3. Hit Record — live transcription starts instantly. When you're done, click Summarize.

That's it. Everything runs on your machine.

Download

⚠️ Important — Read Before First Launch

The app is not code-signed yet, so your OS will show a security warning the first time you open it. This is expected for open-source apps — the full source code is available in this repository.

xattr -cr /Applications/EchoNote.app

chmod +x EchoNote_*.AppImage
./EchoNote_*.AppImage

Why EchoNote?

Most meeting tools send your audio to the cloud. EchoNote doesn't.

Features

Core

• Live transcription — Real-time speech-to-text powered by Whisper, supporting 90+ languages.
• Speaker identification — Automatically detects and labels different speakers using neural embeddings.
• AI summaries — One-click meeting summaries generated by a local LLM. Tokens stream in real-time.
• Conversational chat — Ask follow-up questions about any meeting ("What did Maria say about the deadline?").
• Meeting notes — Take timestamped notes alongside the live transcript; include them in AI context.
• Full-text search — Instantly search across all past meetings.

Customization

• Custom summary templates — Create prompt templates for any format: 1:1s, sprint reviews, sales calls.
• Model selection — Download and switch between multiple ASR and LLM models at runtime — no restart needed.
• Hardware-aware recommendations — EchoNote detects your RAM and suggests the optimal models automatically.
• Theme switcher — Light, dark, and system-follow modes.

Integrations

• MCP server — Expose your meeting transcripts, notes, and summaries to AI coding assistants (VS Code, Cursor, Windsurf, Claude Desktop) via Model Context Protocol. One-click install from Settings.

Reliability

• Works with Teams, Zoom, etc. — Captures audio at the device's native sample rate, so it coexists with apps that use exclusive audio formats.
• Sleep prevention — Automatically prevents OS sleep during recording.
• Auto-updates — The app checks for new versions on launch.
• Setup wizard — Guided onboarding: permissions, hardware check, model downloads, and test recording.

How It Works

🎤 Microphone → Silero VAD → Whisper → Speaker ID → Live Transcript
                                                          ↓
                                              Qwen 3 LLM → Summary / Chat

All processing happens locally using open-source models:

Models are downloaded on-demand from the built-in model manager (Settings → Models).

System Requirements

Privacy & Security

• Zero network access during meetings — all AI runs locally.
• No telemetry. No usage data, analytics, or crash reports.
• Local storage only. Meetings live in a SQLite database on your machine.
• Open source. Audit every line of code in this repository.

Roadmap

• Live transcription with Whisper (90+ languages)
• Speaker identification (diarization)
• Meeting persistence and full-text search
• Local LLM summaries with streaming
• Conversational chat with meeting context
• Custom summary templates
• Runtime model selection
• Timestamped meeting notes
• Hardware-aware model recommendations
• Setup wizard & onboarding
• Light / dark / system themes
• MCP server for AI coding assistants
• System audio capture (transcribe the other side of the call)
• Encrypted local storage (SQLCipher)

Built With

Contributing

Contributions are welcome — see CONTRIBUTING.md for guidelines.

PolyForm Noncommercial 1.0.0 © 2026 Luis MC