Hi everyone!

I wanted a fast, frictionless way to dictate text on Windows without relying on clunky built-in tools or expensive subscription-based software. So I built Aura — a lightweight, native helper written in Rust (Tauri).

It runs in your system tray, listens to a global hotkey (like Alt + V), and transcribes your voice directly into any active text field.

GitHub Repository: https://github.com/malashkadev/aura

https://i.redd.it/z5xmkblmmgbh1.gif

Key Features:

• Global Hotkey & Tray Integration: Hold to talk, or short-press to latch recording (toggle mode). Esc cancels.
• Dual Engines:
  • Local: Fully offline transcription using whisper.cpp (automatically downloads Tiny/Base/Small/Medium models).
  • Cloud: Gemini, OpenAI, or Groq (Whisper + Llama) for near-instant execution.
• AI Cleanup: Automatically formats transcripts, removes filler words (like "um", "uh"), fixes grammar, and structures paragraphs.
• Privacy Focused & No Subscriptions: You use your own free-tier API keys (e.g. Groq/Gemini have massive free tiers) or run 100% locally.
• Focus Guard: Ensures that if you switch active windows while the transcription is processing, the text won't be typed into the wrong window.
• Open Source: Fully licensed under AGPLv3.

Why build this?

Most voice dictation software either sends all your telemetry to unknown servers, locks basic features behind a subscription, or types raw transcription files filled with "uhs" and "ums" without formatting. Aura acts as a system-wide wrapper that pipes your voice to the engine of your choice, cleans it up, and simulates typing instantly.

I'd love to hear your feedback, feature requests, or contributions! Feel free to check out the repo and grab the pre-compiled installer.