Voice dictation is one of the most powerful ways to think, write, and code at the speed of thought. But when I looked at the existing landscape, nearly every tool charged $12 to $20 a month for what was essentially a simple API wrapper bundled inside a 400MB Electron app that routed private microphone audio through centralized servers.
I believed the entire concept could—and should—be completely free. With OpenAI's open-weights Whisper Large v3 Turbo and Groq's high-speed LPUs running inference at ~216x real-time speed, high-speed transcription can run directly on a personal, free API key for zero cost.
As a lone engineer, I spent the last month building Rusper from scratch. I built it natively in Rust using Tauri v2 so it would consume only ~25 MB of RAM, deliver sub-300ms end-to-end latency, preserve your active window focus, and inject text cleanly into any Windows application via native Win32 APIs with zero telemetry.
Project Artifacts
- Source Code: GitHub Repository (AGPL-3.0)
- Windows Installer: Download Latest Release (.exe)
- Knowledge Base: Frequently Asked Questions
- Issue Tracker: Submit a Bug Report or Feedback
How to Contribute
Rusper is fully open source. You can contribute by writing Rust for cross-platform support (macOS/Linux), integrating offline transcription via whisper.cpp, crafting custom prompt personas, submitting phonetic vocabulary mappings, testing edge cases on different hardware, or simply starring the project on GitHub.
Available for Hire
I am an engineer available for full-time roles and contract work. I specialize in building high-performance desktop systems in Rust, low-latency AI inference pipelines, and fast web applications with React and Next.js.
You can explore my portfolio, view my other work on GitHub, or reach me directly at ajinkyaadharkar@gmail.com.
Ajinkya
Creator of Rusper