moinulmoin/voicetypr

What it solves

Voicetypr provides a way to perform voice-to-text dictation across any application on macOS and Windows. It eliminates the need to manually type by transcribing spoken words and automatically inserting the text at the active cursor position.

How it works

The app uses a combination of local and cloud-based transcription engines. By default, it is offline-first, utilizing Whisper and Apple Silicon-optimized Parakeet models for local processing. Users can also opt into cloud providers like OpenAI, Groq, and Deepgram. Once audio is captured via a global shortcut, the system transcribes the audio and can optionally use AI (via providers like Anthropic or Gemini) to format and clean up the rough dictation before inserting it into the target app.

Who it’s for

It is designed for users who want a system-wide dictation tool that works across all software, as well as developers or AI agents who need a programmatic way to capture audio and convert it to text via a CLI.

Highlights

• System-wide integration: Works in any app using global shortcuts and push-to-talk.
• Offline-first: Local transcription ensures privacy and removes per-minute API fees.
• AI Formatting: Integrates with LLMs to refine and clean up transcribed text.
• Agent-ready CLI: Provides a structured JSON output for scripts and local AI agents.
• Cross-platform: Native performance on macOS and Windows using Rust and Tauri.