Talky

A free, open source meeting notes app with live transcription and AI-powered note enhancement.

Talky is a desktop application built with Tauri (Rust + React/TypeScript) that lets you take notes during meetings while automatically transcribing both sides of the conversation. Transcription runs entirely on your device — your audio never leaves your computer. Optional AI features let you enhance your notes and chat with your transcript.

Features

<table> <tr> <td width="50%"> <img src="docs/images/01-take-notes.png" alt="Take your own notes" /> <p><strong>Take your own notes</strong><br/> Jot down thoughts, agenda items, and key points in the built-in editor while Talky records and transcribes in the background.</p> </td> <td width="50%"> <img src="docs/images/04-transcript.png" alt="View the full transcript in real time" /> <p><strong>View the full transcript in real time</strong><br/> Access the complete timestamped transcript as it's generated, with your mic and speaker audio clearly separated.</p> </td> </tr> <tr> <td width="50%"> <img src="docs/images/02-ai-enhance.png" alt="AI-enhanced notes" /> <p><strong>Enhance with AI</strong><br/> One click transforms your rough notes into polished, detailed meeting notes — AI fills in context from the transcript while preserving your original thoughts.</p> </td> <td width="50%"> <img src="docs/images/03-ask-questions.png" alt="Ask questions about your meeting" /> <p><strong>Ask any question</strong><br/> Chat with your meeting in natural language. Ask about decisions, action items, or anything discussed — AI has full context of your notes and transcript.</p> </td> </tr> </table>

Quick Start

Installation

macOS (quick install)

curl -fsSL https://raw.githubusercontent.com/itskhalil/talky/main/install.sh | bash

Windows (quick install)

irm https://raw.githubusercontent.com/itskhalil/talky/main/install.ps1 | iex

Development Setup

For detailed build instructions including platform-specific requirements, see BUILD.md.

How It Works

1. Create a Note — open the app and start a new note
2. Type your own notes — jot down thoughts, agenda items, or context in the built-in editor
3. Start recording — hit record and Talky transcribes both your microphone and system audio (speaker) in real time
4. Stop and resume — pause recording whenever you want, start again within the same note
5. Enhance with AI — use AI to polish your notes, filling in details you missed from the transcript
6. Chat — ask questions about your meeting and get answers based on your notes and transcript

Local Transcription

All transcription happens on your device:

• Mic and speaker audio are captured and transcribed separately so you can tell who said what
• Silence is filtered using VAD (Voice Activity Detection) with Silero
• Transcription uses Parakeet models

AI Features

Talky integrates with AI providers for enhanced productivity:

• Enhanced Notes — AI merges your rough notes with transcript details to create polished, comprehensive meeting notes. Your original notes are preserved and clearly marked alongside AI-extracted content.
• Chat — Ask questions about your meeting in natural language. The AI has full context of your notes and transcript to provide relevant answers.

Supported AI Providers

• Local: Ollama, Apple Intelligence (macOS Apple Silicon)
• Cloud: OpenAI, Anthropic, OpenRouter, Groq, Cerebras (disabled by default, enable via debug pane)
• Custom: Any OpenAI-compatible endpoint

Architecture

Talky is built as a Tauri application combining:

• Frontend: React + TypeScript with Tailwind CSS for the notes and settings UI
• Backend: Rust for system integration, audio processing, and ML inference
• Core Libraries:
  • transcribe-rs: Local speech recognition with Whisper and Parakeet models
  • cpal: Cross-platform audio I/O
  • vad-rs: Voice Activity Detection
  • rubato: Audio resampling

Debug Mode

Talky includes a debug pane for development, troubleshooting, and advanced settings like enabling cloud AI providers. Access it by pressing:

• macOS: Cmd+Shift+D
• Windows: Ctrl+Shift+D

Platform Support

• macOS (Intel and Apple Silicon)
• Windows (x64 and ARM64)

License

MIT License - see LICENSE file for details.

Acknowledgments

• Handy by CJ Pais, for inspiration and the core transcription stack
• Whisper by OpenAI for the speech recognition model
• whisper.cpp and ggml for amazing cross-platform whisper inference/acceleration
• Silero for great lightweight VAD
• Tauri team for the excellent Rust-based app framework