Ciri from The Witcher (game series and R34) is on NAI, IL, PonyXL and SD1.5 animefull now!
Tags:
General:
1girl, ciri, witcher medallion, white hair, green eyes, freckles, updo, single hair bun, scar across eye
Main game appearance:
white shirt, collarbone, arm belt, brown gloves, leather corset, leather corset, brown leather pants, silver belt, black belt, multiple belts, belt pouch, brown footwear, knee boots
DLC appearance:
chainmail, crop top, red sleeves, midriff, pelvic curtain, black pants, black knee boots, knee pads, shoulder belt, fingerless gloves, red thigh boots, studded belt
optional:
holding sword, sword, weapon over shoulder, sheathed, weapon on back
bimbofication, curvy, huge breasts, thighs
scar on leg, scar on stomach, scar on arm
3d
Also for 3D style you can try use this lora with my 3D loras for more 3D effects / 3D styles.
But Ciri lora is also very good in 3D as you can see in my preview.
NoobAI-XL quality prompt:
Put "masterpiece, best quality, good quality" in positive and put "low quality, worst quality" in negative.
Use euler, cfg 4-6. Don't use higher CFG!
If you have over-constrasted picture and cfg 4-5 didn't help you can try to put in negative "colorful".
For NoobAI-XL V-pred:
Check my guide about V-pred image generating!
The best model to use V-pred is Cyberfix! But pure NoobAI-XL V-pred is also good.
For Illustrious mixes and NoobAI-XL Epsilon: use NoobAI-Xl Epsilon (especially mixes based on NoobAI-XL Epsilon!).
Guide about NoobAI-XL Epsilon
PonyDiffusion v6 (XL) quality settings:
Style settings for PonyXL before prompt (important):
score_9, score_8_up, score_7_up, score_6_up, uncensored,
Negative prompt for PonyXL (recommended):
dutch angle, censored, watermark, jpeg artifacts, muscular female
SD1.5 quality settings:
masterpiece, high quality
in negative: worst quality, low quality
Upscaler / adetailer settings:
Hires.fix: Upscaler foolhardy_remacri (download and put in "ESRGAN" folder in "models" if you are using automatic1111 or forge).
Hires steps: 12-16
Denoising strength: 0.2-0.3
Adetailer: denoising strength 0.25-0.3. Detection model confident threshhold: 0.7-0.8.
Best sampling methods (with 25-40 sampling steps for txt2img):
euler a SGM Uniform, CFG 5-6;
euler ancestral CFG++ SGM Uniform / Simple, CFG 1.2-1.5.
You don't need any DPM++.
I will be glad to receive feedback from you and a demonstration of your art with my LoRA!
Description
SD1.5 animefull (clip skip 2)
FAQ
Comments (9)
Great! <3 U amazing !
Cool SFW and NSFW lora, amazing job and thank you so much.
Feel like addillustri https://civitai.com/models/832573
is better then animescreenshot merge, but let me know what you think.
I'll try this model too 😊
I tested a little but it feels similar - https://civitai.com/posts/23585660
Maybe I need more tests.
Since I usually use models with my loras I feel smaller impacts from model so it can be better. Do you have any pics for comparison?
@Enigmata Yeah, like in the photo you posted, I find illustrious is just a little better with color then anime screenshot as I find the latter can wash them out sometimes. Though it's not without its drawbacks it tends to bias game cg in mixed datasets so getting a pure anime screenshot is almost impossible.
@darkspaceinfinity if you want pure anime screenshot you can use NoobAI vpred 1.0 but it will copy all blurry/lowres effects if you will use loras for old hentai characters (like my loras for Bible black) because V-prediction models are better in catching dataset style but it also copies all artifacts.
But this problem can be resolved by using Cyberfix version - it is the same NoobAI Vpred but with swapped SDXL 1.0 on Cyberrealism XL in recipe. It gives better anatomy and style stabilization. So if you will use anime screenshot tag with character loras (or without if model knows char) on cyberfix you will get anime screenshot but remastered (without blurry/lowres effects).
@Enigmata Thanks for the info, I'll definitely try that out. You think it will work on a greyscale manga dataset character?
@darkspaceinfinity I think yes