When she comes packaged with instructions, I think. It's kinda hard to read... This LoRA helps with English/western text body writing. Obviously ai is still pretty garbage at text, but after playing around with a few LoRAs featuring characters with tattoos that it was able to capture rather faithfully, I figured I'd give it a shot. While most checkpoints are pretty good at Japanese shaped body writing and tallys out of the box, English and western style tallys rarely work by prompt alone, hence this LoRA. Again, don't count on the letters actually making sense, but they should at least be English letters, or vaguely shaped like them at any rate. The main activator is b0dy_wr171ng. The variant is for western style tallys. Honestly the main reason I made this LoRA is for that, I'm just not a fan of the squarish Japanese style of tally. This makes tally marks on the skin closer to the 4 vertical lines and one diagonal per 5 marks as is done in the west. Activator is b0dy_wr171ng_74lly. It might not always get the number of vertical lines right, but it should be in that style. Again, ai means it will probably rarely be genuinely legible, but this is more to get the anesthetic and vibe closer to the body writing I like. Strength of 1 works best for me. ALSO:

I Would like to mention some functionality that I tried to train in but seem to have failed at. Note, these other variants don't really work, but if you want to try them, here's how they're meant to work, b0dy_wr171ng_4rr0w: A drawn on arrow pointing to a hole, typically a "cum here" function etc. b0dy_wr171ng_wh0r3: Whore spelled out with the O replaced by a hole: mouth, anus, or pussy. b0dy_wr171ng_l157: A fairly long orderly block of text, like a signup sheet or a pricing menu. b0dy_wr171ng_l4bl3d: The opposite of list, a few large words written on her to label her, like "fuck toy" across her chest for example. Even though they don't actually do what I put them in there for (at least not for me), they still effect the LoRA by skewing which parts of the training data it's looking at. b0dy_wr171ng_4rr0w might fail to draw arrows, but you might like the placement and shapes of the regular writing better with it enabled for example.

As it stands, for those who can draw/write well, I think the best use case for this LoRA is for reference via image to image for properly writing in the text yourself. Use the LoRA to show how the letters curve with the body and are effected by lighting, then go back to the original image and write words in yourself that can actually be read, but with some solid help in figuring out how to most smoothly integrate it into the existing image.