I suspect the issue here was confusion around what “it” was as the sentence has two possible subjects - “the cat” and “tail”. Given cat was the primary subject, it probably assumed you wanted the arrow pointing to the cat.
You didn’t tell it to point at the tail, so it pointed to it (the cat). Task failed successfully.
I said, can you please draw me a cat without a tail, then I asked this one, it has a tail can you draw an arrow arrow to it. They all had tails
ask for a “manx cat”, they don’t have tails ☞ https://en.wikipedia.org/wiki/Manx_cat
It literally said at the bottom within your screenshot ‘Gemini can make mistakes, so double check it’
Is this a common thing with AI images so far, it doesn’t know anatomy, just locations of where to draw said names?
If you want to mess with Omnigen it was designed for this kind of thing. The code and model were released a few days ago.
