Post by squariarch in (Link Fixed.. Preg. hopefully fixed) Minimally Fucky Diffusion Megapack + Mod - ~5k body and portrait sets, ~22k images including clothed, exposed, and pregnancy, male and female every race, hair color, etc. - itch.io

Viewing post in (Link Fixed.. Preg. hopefully fixed) Minimally Fucky Diffusion Megapack + Mod - ~5k body and portrait sets, ~22k images including clothed, exposed, and pregnancy, male and female every race, hair color, etc.

↑ View parent post

squariarch360 days ago (1 edit)

So 1. seems pretty much unsolvable, at least for now. 90% of the work in this pack was trying to convince it to do approximately the right thing IE not add like extra arms and so forth, something like making sure there aren't also human ears seems like it is probably impractical.

Number 2. is my fault largely, and I'll consider it if there is a version 2. I think there is a tradeoff though in the variety/interest of clothing if you try to make everything preindustrial, which I know makes it "realistic" but if ultimately the characters seem more samey because of that I don't know that it is necessarily worth it. But like I said, I'll consider it esp with respect to zippers and other outliers.
Edit: Also yeah definitely if/when I do another version I'll at least partition the outfits more by race so they fit better thematically.

Guileless359 days ago (1 edit)

Thank you for responding!

I want to say that, despite most of my post being critical, I'm impressed with the pack and, no disrespect to Maverik et al, it's the best solution so far if the goal is to comprehensively represent the variety of visible attributes described within the game with aesthetically pleasing, stylistically consistent, imagery.

You employed a new, burgeoning, technology that goodness knows probably would have required exponentially more effort to use 2 years ago and may require exponentially less 2 years from now. I imagine that improving on your result will become increasingly possible in the future but whether you iterate on this project or someone else takes up the gauntlet I doubt that I will be quite as dramatically impressed; you did it first. Still, it's a sign of things to come, it's pretty interesting to think about what will be possible as tools and expertise inevitably improve.

One thing I noticed after making my post, with regards to my first nitpick, was that the portraits and bodies are often slightly different and, in many cases, the portrait versions have additional hair or other features obfuscating redundant, human, ears that are visible on the full body image so that only the kemonomimi are visible. I don't know if you touched some of them up manually or by what other means the improvement was made in those cases but I want to explicitly acknowledge that you were already clearly aware of and actively addressing the issue.