Having two different voice models drives me crazy. One can browse the internet, has an American accent, and speaks like a Latin American (in Spain). Sometimes it speaks in English, and I don’t understand a thing. It takes a long time to transcribe, and when it responds, it gets overly wordy (too much). I’ve tried asking it to be brief, but even when it shortens its responses, they are still too long. Additionally, there’s a frequent error where it somehow interprets that I’m saying something like “follow me on my channel,” or “watch my video and give me a like” (not literally, but something along those lines).
Then there’s the advanced version. It can’t browse the internet, its voice is different—less masculine—without an American accent, although sometimes a bit of an Argentine accent slips in. It can be interrupted while speaking, which I love! But if you enter the chat and try to copy something, the interaction is lost, and you have to start over.
Its most common error is that the internet-browsing voice with the American accent suddenly pops in saying something like, “My programming doesn’t allow me to talk about that.” Once this starts happening, it happens frequently. Also, sometimes it doesn’t finish saying everything, so I have to go to the written chat to read it.
I wish they would merge the two with the best features of each one because if they combine all the flaws, it will be a disaster.
Another thing I’ve noticed is that the more they’re personalized, the more different they become. It’s subtle, but I’ve asked both of them the same questions, and each gave me a different answer.