Offline, privacy-first
speech-to-text for desktop
Press a global shortcut, speak, and paste transcribed text into any app instantly. Powered by local Whisper and Parakeet models.
Everything you need.
Nothing you don't.
Escribbo is built for speed, privacy, and seamless integration into your daily workflow.
100% Offline & Private
No telemetry, no cloud APIs. Your voice never leaves your machine. Perfect for sensitive documents and private conversations.
Keyboard-First
Customizable global shortcuts. Choose between toggle mode, push-to-talk, or trigger post-processing anywhere.
Blazing Fast
Transcribe with Whisper (GPU accelerated) or Parakeet V3 (CPU-only, ~5× real-time) with automatic language detection.
Material You UI
Beautiful native-like interface. Pick a seed color and watch the entire app and recording overlay tint to match your style.
CLI & Automation
Start via terminal flags like --toggle-transcription. Integrate easily into your custom scripts and workflows.
Open Source
MIT Licensed. A robust fork of cjpais/Handy built with Tauri, Rust, React, and Tailwind. Hackable and free forever.
How it works
A frictionless workflow designed to keep your hands on the keyboard.
Press
Hit your configured global shortcut (e.g. Ctrl+Shift+Space) to start listening instantly.
Speak
Talk naturally. A sleek overlay appears to let you know it is recording.
Release
Release the keys (push-to-talk) or press again (toggle) to finish recording.
Paste
Escribbo privately transcribes the audio and simulates keystrokes to type it right where your cursor is.
Get Escribbo
Finding latest release...
Documentation
First Launch
After installing, launch Escribbo. The app will sit in your system tray. On first launch, you may need to grant Microphone permissions and Accessibility permissions (on macOS/Linux for global hotkeys and simulated typing).
Choosing a Model
Escribbo supports multiple backends:
- Whisper (GPU/CPU): Highly accurate, comes in different sizes (Small, Medium, Large, Turbo). Open the settings to download a model.
- Parakeet V3 (CPU-only): Specifically optimized for blazing fast CPU inference (~5× real-time) with automatic language detection.
Verification
Releases are signed using minisign to ensure integrity. The public key is:
RWQ2n+m+S1tTqf5vFq2w7F8Bq9z8u8/K+G+v8w8w8w8w8w8w8w8w8w8=