I'm not seeing the tallness in the example images. If you do another version of this, I would recommend training it to include something in the image that can be used for reference.

It's like in D&D artwork. Bad art of a hill giant is just a dude. Good art is the same dude sitting on a fallen tree like it's a bench.

Not saying that Envy Goddesses are that big, but the same thought process applies.