I do understand what you mean, and I wouldn't be surprised of components of ChatGPT are implemented for a voice command A.I. Artificial intelligence is generally divided into two types – narrow (or weak) AI and general A.I., also known as AGI or strong A.I.[1] ChatGPT is what's known as a narrow A.I., it can only predict what text is should say based on the data it has. The Jarvis you're talking about is a general or strong A.I.
[https://certes.co.uk/types-of-artificial-intelligence-a-deta...]
Eleven Labs has a very realistic British voice that can be relatively easily connected to ChatGPT output. Sometimes the latency isn't great at the moment for the inexpensive plans but I think they are working on it.
But we should expect better models and techniques within the next couple of years and potentially much, much better if you are talking about several years.
The most obvious (in my opinion) upcoming improvements are multimodal large models and further down the line new compute-in-memory substrates with much better efficiency/performance potential.
But also more sophisticated cognitive architectures are being developed and are likely to become more widespread in the next several years.
But you can get text to speech libraries pretty easily to try it for yourself.
It's far smarter than Alexa. What do you mean?