OpenWhispr

Overview

OpenWhispr is an open source, privacy-first voice-to-text app for dictation, meeting transcription, and notes. It can run local speech-to-text models or use cloud processing with bring-your-own API keys, and is available cross-platform for macOS, Windows, Linux, and iOS.

Key Features

• Global-hotkey voice dictation into any app with automatic pasting.
• Local speech-to-text with models such as Whisper and NVIDIA Parakeet, or cloud processing with bring-your-own API keys.
• AI agent and Voice Assistant hotkey for commands, in-place editing of highlighted text, and pasted answers.
• Meeting transcription with auto-detected Zoom, Teams, and FaceTime calls, live speaker diarization, voice fingerprinting, and calendar integration.
• Notes with folders, semantic search, cloud sync, AI actions, and team spaces with sharing controls.
• Audio and video import with batch upload, URL/paste support, and optional speaker detection.

Use Cases

• Dictating into chat, email, docs, and coding apps using a global hotkey.
• Capturing meeting notes without a bot in Zoom, Teams, and FaceTime calls.
• Transcribing existing audio or video files, including batch uploads and URLs.
• Running private offline transcription with local models or using cloud/BYOK processing for speed.

Getting Started

• Clone the repository, run npm install, and run npm run dev.
• Node.js 24+ is required.

Deployment & Requirements

• Available for macOS, Windows, Linux, and iOS.
• Node.js 24+ is required for development.
• The desktop application runs from the repository root; the Expo mobile application has its own dependencies, lockfile, build configuration, and release process.

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.
• Cloud transcription is optional, and recordings are deleted after transcription finishes; local dictation keeps audio on the user’s device.