Stop Typing. Start Dictating.
Dictate code comments, documentation, commit messages, LLM prompts, and emails — 3x faster than typing. Works in VS Code, Cursor, terminals, and every other app on your machine.
View on GitHubWhy Developers Need Better Dictation
You spend hours every day writing things that aren't code: pull request descriptions, documentation, Slack messages, emails to stakeholders, prompts for ChatGPT and Claude. Typing all of that is slow, and your wrists are already taking a beating from actual coding.
Most voice dictation tools are built for general consumers. They don't understand developer workflows, they can't format output as a docstring or code comment, and they send your audio to someone else's cloud — not ideal when you're dictating proprietary architecture decisions.
OpenWhispr was built by developers, for developers. Local processing keeps your IP private, and agent mode transforms raw speech into formatted output — code comments, commit messages, documentation, whatever you need.
How OpenWhispr Fits Your Dev Workflow
Agent Mode — AI-Powered Formatting
Speak naturally, get formatted output. Custom system prompts transform your dictation into code comments, JSDoc blocks, commit messages, or any format you define. Say "this function validates the user input and returns a sanitized object" and get a clean docstring.
Works in Any IDE or Terminal
OpenWhispr works system-wide via clipboard paste simulation. VS Code, Cursor, Neovim in a terminal, JetBrains, Xcode — if you can paste text, OpenWhispr works there. No plugins, no extensions, no configuration per app.
Push-to-Talk Hotkey
Hold a key to dictate, release to transcribe. Or toggle mode if you prefer. The hotkey works globally — you never leave your current context. It's the same muscle memory as push-to-talk in Discord or Slack huddles.
Local Processing — Your Code Stays Private
Every transcription runs locally on your machine via OpenAI Whisper or NVIDIA Parakeet. No audio leaves your device. When you're dictating about proprietary systems, architecture decisions, or client work, your data stays exactly where it should.
Real-World Developer Scenarios
Writing LLM Prompts and AI Chat Messages
You're working in Cursor or ChatGPT and need to describe a complex problem. Instead of carefully typing a multi-paragraph prompt, hold the hotkey and explain what you need out loud. OpenWhispr transcribes it, and agent mode cleans up filler words and structures the output. Your prompts get more detailed because speaking is effortless.
"Refactor the authentication middleware to support both JWT and session-based auth, keeping backward compatibility with the existing API routes"
// OpenWhispr outputs the clean text instantly
Documenting Code As You Go
You just finished a tricky function. Set your agent mode system prompt to "Format as a JSDoc comment" and dictate what the function does. OpenWhispr outputs a properly formatted comment block. Do the same for README sections, architecture decision records, or inline explanations.
/**
* Validates user input against the schema and returns
* a sanitized object with normalized field values.
* Throws ValidationError if required fields are missing.
*/
Commit Messages, PRs, and Slack
Writing good commit messages and PR descriptions takes time. With OpenWhispr, speak a natural description of your changes, and agent mode formats it as a conventional commit message or PR body. Same for Slack updates — hold the hotkey, explain the status, and get clean prose instead of half-formed thoughts.
Why Developers Choose OpenWhispr
Open Source
Inspect the code, audit the privacy claims, contribute features. No black boxes.
Cross-Platform
macOS, Windows, Linux. Same workflow on your MacBook and your Linux dev server.
Free Forever
All core features at zero cost. Optional $8/mo Pro for pre-built binaries and auto-updates.
Your Keyboard Will Thank You
Join thousands of developers who dictate docs, prompts, and messages instead of typing them. Open source, private, free.
No account required · Works offline · Open source forever