Spent a stupid amount of time playing with Stable Diffusion again, yesterday.

-=-=-

I'm learning the best ways to train the thing to draw things like I do, but it's proving somewhat difficult due to my "not quite normal" styles.
For the past couple of days I've been assuming it's likely down to Platdude not looking all that recognisable as a character, so instead I tried training it using a few images of Greenie.

In went such wonders as..

And out came...



...
Well, I think we can honestly say that whatever should be working, probably isn't working all that well.
Right?!
I mean, that's definitely not my art style doing that.
.. or is it?!
Hmmm..

Maybe I need to redraw Greenie in a more "normal" looking manner, in the hope that the result is a bit better.
Either that, or I'm doing all of this wrong.
I'm wondering if the "Textual Inversion" stuff might be using a different base-SD engine than the one that DrawThings is using.