OpenWhispr is a privacy-first, open-source voice-to-text dictation app for desktop: press a hotkey, speak, and polished text appears at your cursor in any application. It runs fully offline with local speech models or connects to cloud providers with your own keys, and it has grown to nearly 10,000 stars on GitHub with active development through 2026.
OpenWhispr is cross-platform across macOS, Windows, and Linux. Local transcription uses models such as Whisper, NVIDIA Parakeet, and Cohere Transcribe — with GPU acceleration via Metal, CUDA, or Vulkan — so audio never has to leave your device; cloud processing with your own API keys is available when you want maximum speed or quality. Beyond raw dictation, the app turns speech into notes and actions, with AI text processing, speaker diarization, semantic search over your notes, a public API, and an MCP server so other AI assistants can work with your transcriptions. The project emphasizes zero data collection and no telemetry. A companion mobile app is developed in the same organization, extending the same local-first philosophy beyond the desktop.
Free. OpenWhispr is open source under the MIT license. Local transcription costs nothing; if you choose cloud models, you pay your provider directly with your own keys.
Writers, developers, and professionals who think faster than they type — especially those who want Wispr Flow-style dictation without giving a closed cloud service access to everything they say.