Shin Jia Counter:Side - IllustriousXL + Animagine XL + Pony
V4 - Designated for stability, consistency and ILXLv2 stable. Its designed for higher requirement GPU at 28 step, euler a, and 1536 resolution.
V3.5EX is a overhaul to add official Black Rose clone outfit. Most tagging remains similar except some unofficial style is removed in favor of other style to increase versatility. Trained from 250 images. RevB is revised version to increase consistency.
v3 IllustriousXL AND AnimagineXL is the extended complete version of v1.5 IllustriousXL, it's designed for higher consistency and increased versatility than EVEN v1.5, with a capability of less prone to ART change, staying to original art as close as possible. V3 is trained from adjusted 248 images instead, with many changes to dataset to increases consistency. V1.5 ILXL is trained with 252 images. v2.3 is trained with additional 27 images compared to 188 of v2.2, making a total of 215 images, majority are photoshopped, redesigned by me. And the result? Worth the effort, I guess. In case you want to see sample image tagging, go look down below. This one is made after my experience with IllustriousXL advancement. At this version setting, it can be ran with low requirement of just 20 step, euler a, CFG 5, which even most potato 4gb vram GPU can run it now.
IllustriousXL version uses same exact setting as v2.3 BUT ITS INCOMPATIBLE WITH ANY XL MODEL INCLUDING PONY, save for IllustriousXL and its variant. Its a product of testing ground after multiple attempts on this IllustriousXL lora making, and since it turns out great, I'll put it here.
PLEASE REDOWNLOAD IF YOU WANT UPDATED VERSION, THANKS!
Historical log of the model:
V4 - IllustriousXL V2 model - Literally had the knowledge of everything previously had, reminiscence of AnimagineXL v1.5 in consistency, while being all the versatility and everything else that previous version had due to further in-training designated for ILXL v2.
V3.5EX-RevB - IllustriousXL 1.1 final model - Literally had the knowledge of everything previously had, with far superior consistency.
V3.5EX - IllustriousXL 1.1 experimental model. - Had the knowledge of Black Rose clone outfit.
V3 - Illustrious XL and AnimagineXL 4 - Huge versatility, very consistent artstyle. (increased consistency, trained on higher resolution images to follow-up with ILXL 1.0)
V2 FLUX - Upgraded flux version. Removed v1 due to its lower tier nature.
V1.5 - Illustrious XL - Huge versatility, very consistent artstyle. (slightly stabilizing the colors.)
V1 - Illustrious XL - Huge versatility, very consistent artstyle.
V2.3 - Animagine XL 3.1 - Huge versatility (Up to animagineXL3.1 limit), very consistent artstyle.
V1 FLUX - Testing ground. FLUX is a non-NSFW model.
V2.2 - Animagine XL 3.1 - Good versatility, consistent artstyle.
V1 Pony - Huge versatility, inconsistent art.
V2 ALT - Animagine XL 3.1 - BAD versatility, consistent artstyle.
v1.5 ALT - Removed, too much weakness.
V1.5 Animagine XL 3 - Inconsistent artstyle, good versatility.
V1 Animagine XL 3 - Removed due to so much weakness.
Shin Jia is a character is from Counter Side game. Also known as "Himeno Yui/ヒメノ・ユイ in JP. Please leave a comment if you like or have any suggestion. Any comments are appreciated, this is a LORA that I have worked on the most, practicing, try and errors to finally produce it at current complete version, v2.3. Working on Animagine XL 3.1 is far harder than Pony models.
Also! feel free to support me at Madafada1991 is creating SDXL lora for AI generative image.
Any of your cent will definitely motivate me to make more LORAs!
It's recommended to run this LORA for IL version with advokat's Illustrious-XL SmoothFT - smoothFT-SOLID | Illustrious Checkpoint | Civitai OR with Ikena's Hassaku XL (Illustrious) - V2.1 - Style | Illustrious Checkpoint | Civitai OR Raelina's Raehoshi illust XL - v4.0 | Illustrious Checkpoint | Civitai. Recommended setting is 20 step, 5 CFG, euler a.
Use strength between 0.7 to 1.
Preview are generated using Raehoshi v4. Upscaled using 4xRemacri
AnimagineXL v3 version is designed for Anim4gin3XL4 and its variant. Recommended setting is 27 step, 5 CFG.