An HVAC controller may be controlled in response to a natural language audio message that is not recognizable by the HVAC controller as a command, where the natural language audio message is translated into a command recognizable by the HVAC controller. The HVAC controller may be a thermostat including a housing that houses a temperature sensor, a microphone, and a controller. The controller may identify a trigger phrase in an audio stream provided by the microphone. In response to identifying the trigger phrase, the controller initiates processing of the audio stream to identify a command following the trigger phrase and to generate a command understandable by the thermostat that instructs the controller to perform the identified command. The controller then executes the generated command understandable by the thermostat. A user may communicate with the thermostat via the microphone of the thermostat and/or a remote device having a microphone.