OpenWispr

OpenWhispr is a voice dictation tool that transcribes speech directly into any app on your computer. It runs on macOS, Windows, and Linux, and works entirely offline using local AI models. For anyone who types a lot, whether drafting emails, chatting in Slack, or writing code in Cursor, it's a faster alternative to the keyboard.

The core pitch is speed and privacy together. Speaking is roughly three times faster than typing, and OpenWhispr doesn't trade your data for that speed. Audio processed locally never leaves your device. Even when using cloud processing, audio isn't stored or logged after transcription.

Key capabilities:

• Local AI models – choose from several Whisper model sizes (Tiny to Turbo) or NVIDIA Parakeet, trading speed for accuracy. No internet needed.
• AI cleanup – voice commands like "clean this up" or "draft an email to Mike" apply light editing or rewriting to your raw dictation.
• Custom dictionary – add names, jargon, and technical terms (medical, legal, domain-specific) that auto-learn from your corrections.
• 100+ languages – auto-detects language and handles mid-conversation switching.
• Universal text injection – works in any app that accepts text input, including VoiceInk-style targets like Gmail, Claude, Cursor, and iMessage.
• Meeting transcription – records and transcribes meetings with structured notes, decisions, and action items.
• MCP integration – available on Pro plans for connecting dictation into more complex workflows.

You can bring your own OpenAI API keys for unlimited cloud transcription at no extra cost beyond what OpenAI charges. The free tier includes 2,000 words per week and five hours of meeting recordings per month. Paid plans add unlimited transcription, device sync, and an agent mode for chatting over your recorded data.

For teams comparing it to tools like Superwhisper, the main differentiator is the combination of fully local processing, an auditable open-source codebase, and cross-platform support. The code is public on GitHub, so the privacy claims are verifiable rather than just policy language.