Skip to main content

Welcome to OpenWhispr

OpenWhispr is a privacy-first voice-to-text dictation desktop app that works anywhere on your system. Dictate in any application with a simple hotkey, and your transcribed text automatically appears at your cursor.

Quick Start

Get up and running in under 5 minutes

Installation

Download and install for your platform

Features

Explore all capabilities and processing modes

Configuration

Customize hotkeys, models, and preferences

What Makes OpenWhispr Different

OpenWhispr gives you complete control over your voice data with three flexible processing modes:

OpenWhispr Cloud

Sign in and transcribe instantly—no API keys needed. Free tier includes 2,000 words/week, Pro offers unlimited transcription.

Bring Your Own Key

Use your own API keys from OpenAI, Groq, Mistral, or AssemblyAI. Full control over provider and model selection.

Local Processing

Keep your voice data completely private. Audio never leaves your device with Whisper or NVIDIA Parakeet models.

Key Features

Customizable hotkey to start/stop dictation from any application. Default is backtick (`) on most platforms, with compound hotkey support like Cmd+Shift+K.
  • macOS: Optional Globe/Fn key toggle for hardware-level triggering
  • Windows: Native push-to-talk support with low-level keyboard hooks
  • Linux: GNOME Wayland integration via D-Bus, X11 support everywhere else
Choose from multiple AI providers for intelligent text processing:
  • OpenAI: GPT-5, GPT-4.1, o-series reasoning models
  • Anthropic: Claude Opus 4.6, Claude Sonnet 4.5, Claude Haiku 4.5
  • Google: Gemini 3.1 Pro, Gemini 3 Flash, Gemini 2.5 Pro/Flash
  • Groq: Ultra-fast inference with Llama and Mixtral models
  • Local: Qwen, LLaMA, Mistral models via llama.cpp
  • Local processing keeps voice data on your device
  • No analytics or telemetry collection
  • Open source code available for review
  • Secure storage for API keys in system keychain
  • Minimal permissions (microphone and accessibility only)
  • Custom Dictionary: Add technical terms, names, and jargon to improve accuracy
  • Auto-learn: Automatically detects corrections and updates dictionary
  • SQLite History: All transcriptions stored locally in searchable database
  • Export Options: Copy, edit, or delete past transcriptions
Works seamlessly on macOS 10.15+, Windows 10+, and modern Linux distributions.
  • macOS: Native AppleScript integration for reliable pasting
  • Windows: Native SendInput API with terminal detection
  • Linux: Multiple package formats (AppImage, deb, rpm, tar.gz, Flatpak)

How It Works

1

Press your hotkey

Activate dictation from anywhere with your customizable global hotkey (default: backtick `)
2

Speak naturally

A draggable panel shows recording status with visual feedback. Speak in any of 58 supported languages.
3

Stop recording

Press the hotkey again (or release in push-to-talk mode) to stop and begin transcription.
4

Text appears

Transcribed text automatically pastes at your cursor location. View full history in the Control Panel.
OpenWhispr supports 58 languages including English, Spanish, French, German, Chinese, Japanese, and more. Set your preferred language or use auto-detect.

Processing Options Explained

OpenWhispr Cloud

The fastest way to get started. Sign in with Google or email, and start transcribing immediately.
  • Free Plan: 2,000 words per week with 7-day Pro trial for new accounts
  • Pro Plan: Unlimited transcriptions at $9/month
  • No Configuration: No API keys or model downloads required
  • Account Features: Email verification, referral program, usage tracking

Bring Your Own Key (BYOK)

Use your own API credentials for maximum flexibility. Supported Providers:
  • OpenAI Whisper API: Industry-leading accuracy with whisper-1 model
  • Groq: Ultra-fast inference, cost-effective for high-volume use
  • Mistral: Voxtral transcription with competitive pricing
  • AssemblyAI: Advanced features like speaker diarization
  • Custom Endpoints: Point to your own API-compatible services
API keys can be set through the Control Panel or .env file. Keys are stored securely in your system’s credential manager.

Local Processing

Complete privacy with on-device transcription. Choose from two engines:

Whisper (via whisper.cpp)

OpenAI’s Whisper model optimized for CPU with GGML quantization:
ModelSizeSpeedQualityUse Case
tiny75MBFastestBasicQuick notes
base142MBFastGoodRecommended balance
small466MBMediumBetterQuality-focused
medium1.5GBSlowHighProfessional work
large3GBSlowestBestMaximum accuracy
turbo1.6GBFastHighSpeed + quality

NVIDIA Parakeet (via sherpa-onnx)

Fast multilingual transcription with INT8 quantization:
  • parakeet-tdt-0.6b-v3: 25 languages, ~680MB, optimized for speed
  • Best for lower-end hardware or speed-critical workflows
  • Efficient CPU inference with ONNX runtime
Local models require disk space (75MB to 3GB) and are downloaded to ~/.cache/openwhispr/ on first use.

Technology Stack

OpenWhispr is built with modern, battle-tested technologies:
  • Frontend: React 19, TypeScript, Tailwind CSS v4
  • Desktop: Electron 36 with context isolation
  • Database: better-sqlite3 for local history
  • Speech-to-Text: whisper.cpp + sherpa-onnx (local), OpenAI API (cloud)
  • Audio: FFmpeg for preprocessing
  • UI Components: shadcn/ui with Radix primitives

Open Source & Community

OpenWhispr is MIT licensed and welcomes contributions:
  • GitHub: OpenWhispr/openwhispr
  • License: MIT (free for personal and commercial use)
  • Version: 1.5.5 (actively maintained)
  • Stars: Trusted by thousands of users worldwide

Report Issues

Found a bug? Let us know on GitHub

Contribute

Help improve OpenWhispr for everyone

Next Steps

Install Now

Download for macOS, Windows, or Linux

Quick Start

First transcription in 5 minutes

Browse Features

Explore all capabilities in depth

Build docs developers (and LLMs) love