They're separate, so one or the other will work. If you're using an OpenAI API key for handling chats, the same API key can be used for Text-to-Speech. But if you have ElevenLabs (additionally) you could prefer to use their higher quality voices instead. Both require an API key in your system variables to read from to work, though.
For the latest versions, the offline TTS plugin hasn't been updated for the latest Unreal Engine (and the quality was lacking with my current implementation) so the latest upcoming version will have the offline TTS plugin removed if it isn't been already.