Difference between Gemini key approach and token approach?

### Description of the bug:

Sorry, I didn't find this repo before and I started to work on a [Next.js version](https://github.com/LivioGama/nextjs-gemini-audio-stream-realtime)

It's not finished yet, but I based it on [this sample](https://github.com/GoogleCloudPlatform/generative-ai/tree/main/gemini/multimodal-live-api/websocket-demo-app) (I still don't get why there are 2 separate github accounts for the same thing) and it's token based.

So if you know, could you nicely explain the difference to me? Also the nextjs app is running the backend, but what about yours where is it? Can we chat (dialogue) with your project ? I didn't manage to in my trial

### Actual vs expected behavior:

_No response_

### Any other information you'd like to share?

_No response_