You have to prompt it first, e.g. point the camera at a street sign and ask ‘Where am I’, or show it some text and ask ‘what does this mean’.

If you just want direct translations Google Translate is the app you’re looking for, because you point the camera at foreign text and the translated text appears.

Gemini will provide more context. For example if you use Google Translate and point it at a menu item like ‘pierogi’ you’ll get ‘dumplings’. Gemini will say something like ‘this is pierogi, an eastern european dish of dumplings commonly stuffed with meat etc.’

Likewise if you point it at a street sign or landmark you can ask ‘tell me about this place’, ‘give me directions to abc’, etc.

ChatGPT had this feature long before Gemini… you also start it by tapping the button next to the microphone icon.