Saw some other posters expressing this view who deleted their posts after getting downvoted, lol.
They changed the voice to intone like Scarlett Johansson's character. It's like they changed the song the voice was singing to one that lots of people recognise.
What's likely different is that GPT-4o can output the tonality instructions for text to speech now.
It's probably the same voice, but different instructions for generations. One was without tonal indicators, one with.