OpenWhispr is a voice-to-text dictation application that runs on desktop platforms with privacy-first design and support for both local and cloud-based speech recognition models.
The tool addresses the need for private, on-device speech transcription by letting users press a hotkey to dictate text that appears at their cursor. It offers a choice between fully offline transcription using local models like Whisper and NVIDIA Parakeet, where audio never leaves the device, or cloud processing through services like OpenAI, Anthropic, Gemini, and Groq for faster results. Users bring their own API keys for cloud services, maintaining control over their data and avoiding vendor lock-in. The application includes features for meeting transcription and note-taking alongside basic dictation.
Developers choosing this tool should consider it if privacy is a primary concern and they want to avoid sending audio to third-party services. It suits users across macOS, Windows, and Linux who need a lightweight dictation solution without telemetry or data collection. The README positions it as an open-source alternative to WisprFlow and Granola. One limitation on Intel Macs is that live speaker identification and voice fingerprinting are unavailable due to ONNX Runtime compatibility, though meeting recording and transcription continue to work with keyword-based search instead of semantic search.
The project maintains active development with regular releases and a documented changelog. The codebase is written in JavaScript and distributed across multiple platforms with native installers for each operating system. Documentation is available through both a website and API reference, indicating ongoing investment in user guidance and extensibility.
