Pony of aussie Josh Duckworth - a beast of a fitness model and uh, content creator ..
I wanted to try my new training process on readily available high quality images with some tuned refined image treatment processes I've honed over time. This time I wanted to quantify not only input quality, but output quality on a few factors such as matching, CN conformity, prompt adherence and stacked LoRA stability.
Since I found a site that measures likeness to a fairly impressive degree. I know there others out there that do this, but I found this one to be pretty spot on and fast. Using this to make sure my input data conforms to a high degree of likeness, then lo and behold so does the output of the LoRA! One of my quality checks has moved from a subjective me eyeballin' it to now measuring it scientifically to see which epoch I should go with. I even score carded it that I'll use with my future generations (at least ones that I am looking to reproduce someone's likeness).
anywhoo.. 'nuff nerd stuff and back to the main point why you're here: trigger word is joshduckworth
enjoy responsibly..
Right under CivitAI's 10,000 step limit but with over 300 images in the dataset, it took quite some time to bake but I think it worked out well!
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights. Click to Claim
