<h3 id="disclaimer">Disclaimer</h3>All showcase images with the exception of the cover image are straight Txt2Img. The first image was achieved using inpainting and txt2img.If you are interested in a specific character, I recommend you check out their dedicated character LoRA's because they will give you far better results. This was created in order to assist with group compositions. It does fairly well for a first draft. All showcase images are unedited straight from TXT2IMG.The solo character images in this description and the showcase examples are what I got with a single attempt. While not as good as my single character LoRA the accuracy was much better than I expected for the first attempt. There are not as many outfits however in the group LoRA. Only expect 2-3 per character maximum in this one. I did cherry pick the group images after a batch run of 10 images, so don't expect perfect accuracy when prompting multi-character. There is blending between some character combinations and outfits.This LoRA is best with solo / yuri / hetero compositions. From testing, 3girls+ is very unreliable with most of the combinations. Also, Exstia / Spica also tend to overpower one of the other characters. Plenty of room for future improvements.<h3 id="description">Description</h3>For prompting two characters, use format inoino,[character],[outfit],[character2],[outfit2],{prompt}. I have included the main character triggers and their outfit trigger. I added the supplemental strings for the main outfits as well, which may help guide your prompt. For the most part though, this not a LoRA you can just drop into your prompt and get great results. If you want something easier to work with, then use my other LoRA dedicated to single characters.This is a megapack that contains the following character's from the Exs-Tia series:<ul><li>Genderbent Souma (Spica / Shiori) <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9cef6ad8-0ef9-4b28-9e80-d460506c3d33/original=true/9cef6ad8-0ef9-4b28-9e80-d460506c3d33.jpeg" /> </li><li>Rieszverne<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a18d1496-0501-4741-95df-40aa42425929/original=true/a18d1496-0501-4741-95df-40aa42425929.jpeg" /></li><li>Yuri<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/89d1b4b7-04fc-4a64-bc6b-276f43ab6558/original=true/89d1b4b7-04fc-4a64-bc6b-276f43ab6558.jpeg" /></li><li>Marina<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a21683f1-23e6-4667-b147-b5848d5c9c8e/original=true/a21683f1-23e6-4667-b147-b5848d5c9c8e.jpeg" /></li><li>Azarin<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/eab7cf28-1ab1-4715-be56-29606e22bcf4/original=true/eab7cf28-1ab1-4715-be56-29606e22bcf4.jpeg" /></li></ul>

The Exs-Tia Megapack (5 characters)

Changed dataset tagging to couple characteristics and outfits to just the character &amp; form trigger words. Outfit tags are included in the trigger words. Slight improvements when prompting multiple characters together, but not significantly better than V2. Biggest change is you should be able to just use the character tags without all the clothes tags. Recommended to use the clothes tags when doing solo prompts or inpainting.

Files with SHA256: 1ef6f585900713fd3fc461a0ba0fa48db8ad6cf487bf939d8a36dc745b649e24