Grok just created a "spicy" video which will give serpentophobes nightmares. It is sureal. If you don't want to see it, don't look at the replies to this post. I will post it there.
Grok has one hell of an imagination. I uploaded a picture of two loons on a lake with some red flowers in the foreground. Grok imagined a humingbird hovering over the flowers and then eating a loon!
After many tries with all sorts of weird errors like babies with 2 heads and babies that are part of her belly and other horrors, I believe Grok has it. The order and degree to which you describe things really matters and deleting past bad attempts matters too.
Grok got the right number of arms on the woman, and maybe the baby's other foot is behind her right elbow? And the woman seems to realize that supporting her babies head is more important than picking the peach, which is cool.