Hi there, this is ToxicHolyGrenade here. Overtime, I am starting to get more and more frustrated with the inaccuracies happening with ChatGPT’s art generation feature. These inaccuracies occur far too often, and don’t always meet my expectations. For example, while I know you guys might not be familiar with TF2 Freaks, I tried to generate art showing Windy Sally Parker/Counteractor, but the appearance sometimes does not look like her, and also too often, she will come out as a male instead of being female.
Similarly, I tried to do Spectra Lumiere, a Firefly Inkling OC to SMG4, but her shorts are not dolphin shorts at all, and instead are longer shorts. The point to all this, is that I have seen AI art generation to better than this, and inaccuracies don’t occur that much. The ultimate question is this; does ChatGPT have a downgraded AI art generation tool purposefully? There needs to be transparency here. If that is the case, then not only is it a con, it betrays the other impressive features that was brought into the table in this recent year. I am open to be proven wrong, if it is a matter of going for the more costly subscription, then maybe the issues won’t happen. But I have not seen any of that. Besides, even if that’s the premise, it is still a problem, because there are some people who are not economically stable. I’d rather have the AI generation feature being overhauled and trained much more in order for the inaccuracies to be vastly reduced. Or at the very most, have the AI to obtain a feature to generate many different versions of the same image, in case inaccurate generations might occur.
Thanks for taking the time to read.
Yours sincerely,
Andrew Paul Matchett/ToxicHolyGrenade.