ユーザーLoRA
XL
ユーザーLoRA
floSR, 1girl, dog girl, furry female, animal ears, animal nose, tail, blue eyes, blonde hair, short hair, maid, maid headdress, maid apron, frills, red bow, ribbon, puffy sleeves, thighhighs, leg ribbon, sandals,
"Bonjour~"Flo (or Fromage Financier) is the quest broker from the Nintendo DS action RPG "Solatorobo: Red the Hunter." The 2nd game from the "Little Tail Bronx" series.PonyXL:I didn't make a single change in the dataset :v) Despite this, it works very well. Gives infinitely more consistent results over the SD 1.5 version. Uses the same prompts and triggerwords below.Has a tendency to put her hands down together between her legs by default while sometimes holding a book. Much of the dataset has her in this pose. If you don't want this, adding "hand between legs" and/or "book" in the negative prompt should easily fix this.SD 1.5:Trained on Indigo V105, 512x768, with 58 images as 2320 steps. Had to dig very deep to find good enough images to work with.Weight: 1Triggerword: floSRUseful tags: 1girl, dog girl, furry female, animal ears, animal nose, tail, blue eyes, blonde hair, short hairOutfit: maid, maid headdress, maid apron, frills, red bow, ribbon, puffy sleeves, thighhighs, leg ribbon, sandalsI plan on bringing most of my other LoRAs to PDXL. I did a test and trained several of my existing datasets and this was the only one which turned out good. The others I tested had one issue or another and will need some significant changes to work consistently. This will likely be my main focus for a bit.I also still plan on making LoRAs for every significant character in this franchise. I hope to have versions for both 1.5 and PDXL if possible. Once I make a Pony version for some of my other existing LoRAs, I hope to finally make some new character LoRAs. Running everything on a fairly weak computer makes everything take forever though, which is why I got burnt out for several months. Images with SDXL take even longer with each image taking several minutes without hires fix.For Pony LoRAs, training at 1024 is just a bit too much for my hardware, so I have to use Civit's LoRA creator. It works well, but costs buzz. I thankfully amassed enough of it to be able to do everything I currently want to without it costing me anything, so it shouldn't be an issue. I still barely know how to use or train on XL, but I'll learn it eventually.Feedback and criticism are always appreciated, be it positive or negative.