Trigger: 4s4hina
Prompts: white dress, bare arms, brown hair, tanned skin
Description
FAQ
Comments (7)
Thanks for the model, and brace for nuclear impact.
well, your lora works even without prompts. if you want, you could train it more to get to the point without much prompting.
and what was the ai thinking in this last image with the ghost arm? i know that i thought something similar while those images got generated, but i did never think that the ai would actually really do exactly that.....
i'm still learning how to train my loras properly. this one i barely described the images, idk if i that was the reason but this one came pretty good
and what ghost arm? i can't find it
EDIT: found it lol
just a suggestion, because generated images had the best effect on them this way:
<lora:4s4hina, white dress, bare arms, brown hair, tanned skin (from abenobashi):1> 4s4hina, white dress, bare arms, brown hair, tanned skin (((drawn by tsuruta kenji, 2000s_(style) )))
load the lora at first, add the prompts and the style of the original artist and the year. danbooru has them all in its database and sd uses those tags...
tl;dr.: use my posted prompts for even better images close to the original
EDIT: for more authenticity use "slender face" as prompt
please take a look at my last uploads. i did not use any detail-tweaker-lora or any extra trigger.
they look absolutely awesome!
if you want, feel free to copy these positive and negative prompts to reproduce it. i would love to share my selfmade prompts to get them known better to the community. but be very careful! they are heavy weight! because of this, my standard prompt for every character model became
(enjoying solo alone time in nature).
without this prompt (i really tested tons of prompts and only this one worked exactly like i wanted images to look like. and it only works with brackets) you will get heavily deformed results or multi-panel images or other horror you don't want to see.
and before someone tries to say it, you can change "nature" to anything else. after this one prompt, you can add other things, but it is trial-and-error, because not everything goes the way one wants to have it.
p.s.: if you are annoyed by that always "looking at viewer"-feeling in your results, just remove "cinematic-accurate". but that makes scenes simply boring, because characters just stand in front of you and do nothing.
EDIT: (enjoying some self-(censored) alone time at home) PLUS in the Negative "unnecessary monochrome" gives you some really nice results. mostly mixed original screenshots used while making this model plus some (censored) moments.
dude, they look incredible, congrats! i'll test this prompt later and post the results here
@nullcamp901 i mean the other prompts i used, too...
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.















