TL;DR:
use the lora at strength 0.65-0.75
Sample images are not cherry picked, that's sort of what you can expect with minimal effort danbooru tagging
any more will work super well or super poorly depending on the seed and prompting
CFG: 6.5
Model: PonyRealism
UI: Fooocus
Sampler: DPM++sde Karras
Steps: 45
Prompt tips: add eve14 to prompt, any other danbooru tags for good feet in positive, foot focus is a good add. This lora plays well with natural prompting as well as danbooru, as I used gpt4o to caption all images before adding the booru tags to the .txt training files.
This lora does work super well with the seated soles LoRA at 0.5 (https://civitai.com/models/263039/seated-pose-focus-on-soles-of-feet). His LoRA does toe count quite well, and my own does composition and texture super well, so its a beautiful combo.
Resolutions that work:
most should be okay but these two are the ones I tried in the 10 minutes i played with the lora before posting
896x1152
768x1280
Description
Not as refined as id like. May upload a better version based on a more diverse dataset if this one gets any traction. Still having around 50% of generations with 4 toes on a foot, but i think i know how i can remedy this in future versions. This lora does work super well with the seated soles LoRA at 0.5 (https://civitai.com/models/263039/seated-pose-focus-on-soles-of-feet).
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.