# Voice input You can use your microphone to control the agent. Simply click the microphone icon, close to the send button, and start dictating. ## Browser considerations Note the voice input is fully controlled by your browser, which implies: * The browser uses your default microphone. You can change which microphone to use either in your browser settings or in your Operating System. * The browser uses your default language. You can configure your default language in your browser settings. * The browser may use remote services for speech recognition. Some browsers (typically Chromium based browsers like Chrome, Edge, etc.) offer the ability to perform on-device recognition for certain languages. You can choose this option in Tidewave settings (see below). * Not all browsers have speech recognition available. In such cases, you will see an error message such as "Voice input failed: network error. Try enabling on-device recognition in Settings". ## On-device recognition For browsers that do not provide remote speech recognition services or for privacy reasons, you may opt into on-device recognition, as long as your browser supports it. To enable it, open up Tidewave Settings and find the "Voice input recognition" label: Voice input settings And choose "On-device" instead of "Default". Once selected, the relevant language pack will be installed locally. If your browser does not support "On-device" recognition, the option will be disabled. ## Tips Here are some useful tips for making the best of voice input: * While using the voice input, you can still use the [Inspector](inspector.md) or click on lines during [Code Review](code_review.md), and Tidewave should correctly inject them as part of your speech. However, some browsers require you to briefly pause before clicking. * Use `Ctrl+Shift+M` (or `Cmd+Shift+M` on macOS) to quickly toggle voice input.