I just wanted to share this technique I came across that I haven't heard anyone talk about.
I came across this video and am using its workflow to create a Pony LoRA for which I didn't have a lot of training images:
I think it is primarily useful for images created by the "Background" section. I just turned off all the sections but the first section and the Background section and deactivated the emotion input nodes to the Background section (as shown in the video) and made a bunch of prompts with different scenarios.
I then plan on using HW Tagger to edit the tags and putting the files into Kohya_ss.