malashkadev/aura

Aura is a free, open source voice dictation app for Windows. The idea is simple: you hold down a hotkey, speak, and release it, and your words appear as clean, typed text in whatever application you were using, whether that is a chat app, a code editor, or a document. It is not just a transcriber though. Aura also cleans up what you said, removing filler words and fixing punctuation, and it can even rewrite text you have already selected based on a spoken instruction. You get to choose how the speech is turned into text. Cloud mode sends your audio to a provider like Gemini, OpenAI, or Groq using your own API key, while local mode runs entirely offline on your own computer using Whisper or NVIDIA's Parakeet model, so nothing leaves your machine. The app also supports spoken punctuation commands, a history of your last fifty dictations, a custom dictionary for names and specific terms it should recognize, and support for eleven languages. The README positions Aura as a free alternative to a paid product called Wispr Flow, and compares itself honestly against a similar free tool called Handy, noting where each one is stronger. Aura's main selling point is the AI cleanup layer combined with the option of either cloud or fully local, private processing. Installing it is as simple as downloading an installer from the project's releases page and running it, though Windows will show a security warning the first time since the installer is not code signed yet. Building it from source requires Rust, Node.js, and WebView2, and uses standard npm commands to run it in development or produce an installer. The project is written primarily in Rust, using Tauri to build the desktop app, and is licensed under AGPL-3.0. It currently has 1 star on GitHub, and the author is actively working on bringing macOS support to the same codebase.