yeah, it's getting so good now that it's indistinguishable. Not the example above, I found that pretty lackluster myself, but that's from being an insider nerd dev who looks at these things all day long. From an average consumer/user point of view, I'm sure it sounds near perfect. You know how movie VFX has been photorealistic for 20 years now. Except it hasn't really, it's just the best in class results at all points. But with the voice and personality demonistration from OpenAI with their latest model, the only thing holding it back now is latency. And that's just a scale problem. It shouldn't be a problem soon.