Updated<h2 id="tldr:-you-should-upgrade-to-v2-unless-you-prefer-v1s-more-stylistic-drawn-results.-2pvf3clzq">TLDR: You should upgrade to V2, unless you prefer V1s more stylistic drawn results.</h2>Turns out it was possible to make this one a lot better, I just got impatient when I released V1.For all intents and purposes v2 will be a lot more consistent with the face, and much higher quality with photoreal models. Showcase images has example prompts to bring out the most out of them if you're struggling.V2 is however more susceptible to breaking under the weight of too many negatives/high vector TI embeds if you're on an anime checkpoint; upping the weight of the j4da_v2 embed and adding (dark brown skin:1.5) should help fix that.<hr />Meet Jada Nobody. Smart, driven, and with curves that can't be ignored.A fairly straightforward new girl to meet; one woman, one file, plenty of potentialDespite a 50/50 balanced data set, this one performs best in 2.5D.This one tends to heavily favor skimpy to no clothes, so be sure to prompt those in.Beyond that, have fun! She's not exhaustively tested yet, I'll be doing more of them in the coming days, and may see if additional training gives a better final result.This one should avoid overlong negative prompts, as well.

Jada Nobody

500 steps, 24 image dataset. Trained on SD1.5, 6 vectors on the embed.This one may get revisited in the next day or two as I keep testing.

Files with SHA256: 488dbc4dd5e9e6e3d9be9190bb231b929e55d0f0beba46f36824e03b913bd1ae