It works perfect on my chips laptop, I choose to test the performance of the model on my --lowvram.
IF you have something better than me, please help me report a bug to other Lora.

I'm a Designer, I only read these things online, not a Coder but I'm still there for you <3

My ielts so bad, sorry, if it confuses you.

I have over 3000 tested on Pixai, checking on:

https://pixai.art/model/1650966455826619025

Checking on Seaart tester:

https://www.seaart.ai/models/detail/17be4966fafa2f95fca05530f51b21b4

sampling step: over>20,

Method: Euler A for random, 2M for anime, SDE for both real&anime style,

Upscaler: Latent, R-ESRGAN 4x+

Denoising: 0.5 ~ 0.7,

CFG: 7 ~ 8.

vae-ft-mse-840000-ema-pruned (has baked in model)

-Recommend quality prompt: (masterpiece:1.2), (best quality),

use between: "realistic/photorealistic" or "anime style", just both them with some lora from @numeratic, @kozue

U can use real person (weight 0.4~0.6) + lora@numeratic (0.6~0.7) + some promt = cosplayer, hav fun.

-Recommend negative prompt: ((worst quality, low quality:1.4, monochrome, blurry:1.4, large head, extra digits:1.4)),

U can use: EasyNegativeV2, ng_deepnegative_v1_75t,

-Recommend "scenery" prompt:

ultra wide-shot, landscape photo, cityscape photo, food photo, seascape, space,

-Quality prompts: masterpiece, best quality, high details, ultra highres, absurdres,

-Negative prompts: (((humans, people, person, 1girl, horse, animal))),

With "scenery" prompts, u can add "national geographic". Make your photo look like mastery.

With "ultra wide-shot" prompts, can make bend-horizon effect, like "wide-angle"on camera.

With "landscape photo", "cityscape photo", "food photo", "seascape", its just another photo "no human".

With "masterpiece, best quality", Why not is "(masterpiece:1.2), (best quality)"? I don know, but with "scenery" highRes, u don't need to add this in your prompt, with: (masterpiece:1.2), (best quality) , can make "horse error" in your img.

Petah the horse is here! - I don know, I think it has something to do with Chinese horse scenery paintings, but this is only a small percentage, about 1~2 / 10 img XD.

-The things my Model prioritizes:

Grey background, (That's what I like most about this model).

Aqua eyes, (If u don hav "Lora" or hav "color + eyes" prompts, the eye color will always be "aqua eyes", use prompts "beautiful eyes & detailed eyes" can make eyes be unique.

NSFW clothes, upper body, auto blur background for u, (u can add prompts "NSFW", make your img look more "sexy" :D , If you need seriousness, just add "NSFW" in your neg.

Real hair, real nose, real lips, real body (don use anime anatomy, u can add midriff, ribs, stomach, pelvic, armpits, breast).

Anime eyes, anime shading hair, anime shiny skin, anime background, anime clound & sky, anime plants.

-Weakness of Model:

Sometimes there are problems with "full body" photos.

No light particles, halo body, lights ray, fog, like RevAnimate or Camelian, light particles can destroy your --lowvram GPU.

It is impossible to create the exact hand of a real person, U can add negative prompts here: (((extra fingers, deformed hands, polydactyl:1.5))) or use EnvyBetterHand Locon.

Cant make rigid building, but works well with anime landscapes.

Cant make "loli" body, it's really difficult, you really need a lot of steps.

Be careful when using "pink eyes/red eyes", "red eyes" is a camera term, and also a disease, so be careful.

Extremely difficult to create facial features, Sometimes the faces are almost the same.

Good luck.

If you need more CamelSus examples, visit me on:

https://pixai.art/@84-tutinthitutu

Megre checkpoint: like a cake, do it until the "ancestor" tells you to stop.

DarkSushi25_v40 (0.65) + Camelianmix25_v2 (0.35) = camelSushi25_v1_03

CamelSushi25d_v1_03 (0.6) + CamelianNSFW25_v11 (0.4) = camelSushi25_v1_04

CamelSushi25_v1_04 (0.6) + ChilloutMix_niprunFP16 (0.4) = ChillCamelSus_v1_05

camelSushi25D_v1_04 (0.7) + MaineUnreal (0.3) = camelSus25D_v1_07

camelSus25D_v1_07 (0.8) + ChillCamelSus_v1_05 (0.2) = camelSus25D_v1_08