In my recent post I mentioned a new (beta) voice assistant device for the use in your Home Assistant system. There are others available on AliExpress based on ESP32 etc. but this one is "the official". More information here.
Today it was delivered. Yes, it is not the sexiest device on earth, it attracts dust as crazy and I absolutley want to hide it (and you could because mics and speakers are very good and it will hear you from afar).
I ordered from a local distributor in Germany and it only took 2 days to arrive on my desk. Setting it up was very easy, you can ask ChatGPT for instructions or watch the video linked on the web page. You only need the OpenAI integration, an OpenAI API key and a paid subscription to OpenAI.
Currently I am using the "gpt-4o-mini" LLM which works fine is one of their cheapest to use.
I tested various commands which are definitely not pre-defined in my system such as phrases like "I want some warm colors in my living room, please set the lights accordingly". This is what it is all about: use natural language instead of repeating sentences to call Alexa routines (which also work great with HA, too). What could I say: it just works! 
There are 3 wake words to configure, I am using the default which is "Okay, Nabu".
I also learned that you can tweak how the system should response. In HA settings for the voice assistant pipeline you can define the "tone" by editing the text given in this property (besides that you also can change the model being used).
As you can see: I changed it to a "Jedi knight" style, so sometimes the answers contain something like "I hope this setting will enlighten you, young Padawan!" - and this is awesome!
I also tried to integrate Google Gemini - it is exactly the same as with ChatGPT. Obtain an API key, add the integration in HA, use the API key and set up a new Voice Assistant pipeline to use - done. It doesn't work as precisely as ChatGPT though, but it's currently fully free, so this is a plus.