There's an interesting comparison in the creative/literary end of llm output too, they're in my experience, dreadful with anatomy. Like, it knows humans have hands, heads, etc, but often times a seemingly limited concept of how anything is connected, or degrees of freedom. (e.g., Why yes, certainly there are many examples of humans rotating their torsos 180º at the hip, seems perfectly cromulent)
I honestly don't know if an image model would help, or if it might analyze the output and go "13 fingers? ship it!" anyway.
The most potato i've run it on was a CB131 chromebook, (which was about a hundred bucks in 2017) and even then, even though video was chugging at 20fps, the main update loop was locked at 60TPS. very impressive. (and helpful for a 9 hour layover on christmas eve, anecdotally)
Does anyone else remember Frets On Fire? That could be a fun mode to try to adapt.