1. Install the native speech extension
Open Extensions in VS Code, find the official VS Code Speech extension from Microsoft, and install it. Language components download when you first select or use a supported speech language.
Use the native extension first because it understands VS Code’s own editor and chat surfaces without relying on macOS Accessibility insertion.
- Install VS Code Speech from the Visual Studio Marketplace.
- Allow microphone access when macOS asks.
- Set accessibility.voice.speechLanguage if the automatic choice is wrong.
2. Choose editor dictation or voice chat
Press Option-Command-V to start dictation in the editor and Escape to stop. Holding the shortcut enables walkie-talkie behavior until you release the keys.
Press Command-I to open voice chat. VS Code can submit the prompt after a pause; set accessibility.voice.speechTimeout to 0 if you want to prevent automatic submission.
3. Understand the native privacy boundary
Microsoft states that VS Code voice support does not require an online connection and that recordings are computed locally rather than sent to an online service.
That statement covers the speech-recognition stage. If you use voice chat with an AI provider, the resulting text and code context still follow the selected chat provider’s data terms.
4. Add IraVoice only for the cross-app or brief workflow
Use IraVoice Dictate when you want one held-key habit across email, notes, browser fields, and other supported Mac text surfaces. Focused fields in Electron apps can still expose insertion differently, so the clipboard fallback remains important.
Use Voice-to-Spec when spoken requirements should become ordinary Markdown with a goal, constraints, and acceptance criteria. IraVoice can hand that brief directly only to its explicit Codex or Claude Code destinations—not to VS Code itself.