Models! Pied!
... that's it, that's the post

Multiple passes through Flux with manual editing in between. Big shoutout to messg for his WAM-combined lora, which I used here along with a couple others.

I've tried pies before, but this is the first time I've gotten them close to what I want. But still a bit rough.

Great efforts. You're making good progress.
You've just reminded me I never released my Pie Lora. One of many things that got pushed back as I worked on a larger Finetune.
Now, the question is, do I release it as is or wait until the next finetune training completes.

Another update and some work in progress. I trained a quick lora using just the 10 images I posted before. Quick and dirty, not good for text to image prompts, but useful for the image and controlnet flows I like.

Still seeing what I can use this for and how, examples attached, little to no editing in these so they are rough around the edges. It's absurd how easy it is to make Flux do decent work with not much custom training.