Research/Command Mode: Voice Control for Power Users
productJuly 202510 min read

Command Mode: Voice Control for Power Users

Deep dive into how we built natural language command recognition for desktop productivity.

Beyond Dictation

Voice-to-text is powerful, but what if you could also control your computer with your voice? Command Mode transforms Whisp from a dictation tool into a complete voice interface for your desktop, letting you navigate, edit, and control applications without touching the keyboard.

Natural Language Commands

Unlike traditional voice control systems that require memorizing specific phrases, Command Mode understands natural language. Say "delete the last paragraph" or "remove everything after the comma" and Whisp understands your intent, even if you don't use the exact expected phrasing.

  • Navigation: "Go to the beginning," "Move down three lines," "Jump to the word 'conclusion'"
  • Editing: "Delete that," "Undo," "Select the last sentence," "Make this bold"
  • Formatting: "Add a bullet point," "Make this a heading," "Insert a link"
  • Application Control: "Open Slack," "Switch to Chrome," "New document"

The Technology Behind Command Mode

Command Mode uses a specialized intent recognition model trained on thousands of examples of natural language commands. The model distinguishes between dictation (content you want transcribed) and commands (actions you want performed), with high accuracy even for ambiguous phrases.

Context Awareness

Command Mode understands context. "Delete that" knows what "that" refers to based on your recent actions. "Make it louder" works when you're in a media player but is transcribed as text when you're writing an email. This context awareness makes voice control feel natural rather than robotic.

Customization

Power users can extend Command Mode with custom commands. Define your own phrases and map them to keyboard shortcuts, text snippets, or application actions. Create workflow-specific vocabularies that make sense for your work.

Activating Command Mode

Command Mode can be activated in three ways: say "Command" before your instruction, hold a modifier key while speaking, or enable always-on command detection for hands-free operation. The choice is yours based on your workflow preferences.

Privacy Considerations

Command recognition happens locally on your device by default. Your commands are never sent to our servers unless you explicitly enable cloud processing for enhanced accuracy. This ensures your voice commands remain private.