Thank you very much, I worked hard for that compliment. My quick answer is complicated. But in general, I (mostly) tag and caption manually. In general, I try not to over tag. Lots and lots of experimenting though.

The only exception to manually tagging that I can think of was Civgran, since they (Civitai) shared the dataset, I reworked the images in a Image2Image batch in ForgeUI and (if I recall correctly) left the tags intact. The Big Jim models were another exception (sort of) I tried to use taggui.exe and (fruitlessly) tried to use it's autotagging feature and tags/.txt files from ChatGPT, which constantly gave me the wrong advice (to be honest it was early days) so I ended up using taggui to add my own tags to it. Taggui was useful for automatically generating the .txt files and editing them in bulk. Everyday there's a new app or node, so there may be better ways that I'm not aware of.

For the past few months, I have been experimenting with ComfyUI and have not quite mastered it.