Cover images are raw output at 1MP resolution. No upscale, no fixes, no negative prompt. Styles are from the base model itself, triggered by prompt.

(1/10/2026): cknb02 v0.302a

Dropped 20% images in dataset that are too messy/complicated, should have less artifacts in background. Also trained with 50% less steps, previous version seems a little bit overcooked.

NOTE:

"a" means it has weak default anime style.

Base model is ChenkinNoob-XL. Which is trained from NoobAI v1.1 with latest dataset.

(10/21/2025): Time to look forward...

SDXL is just too ... old.

If you also want to try some "new" anime models, check out Lumina 2.

• Neta Lumina: Anime-style finetuned model from base Lumina 2 model, trained on danbooru and e621 dataset.

• NetaYume Lumina: Further finetune from Neta with newest dataset.

• RDBT: My finetuned model from NetaYume.

If you just heard about Lumina 2, quick info:

• Released at Jan 2025. Open source. Apache License 2.0.

• DiT architecture (same arch as Flux.1 etc.), strong performance.

• Small and efficient. Only 2B parameters, same as SDXL. (Full bf16 model 5GB in size).

• Use the Flux.1 16ch VAE (better than SDXL 4ch VAE).

• Use Google Gemma 2 2b as text encoder (fully functional chat LLM). Much much better than good-old CLIPs.

You can easily prompt multiple characters like this. see prompt,

Also, I don't think Civitai will support Lumina 2. I've also uploaded the model to TensorArt. You can find me at here. And try Lumina 2 models online. If you don't have a local setup.

Stabilizer

Name is misleading. It's not a tool nor style LoRA.

It is a finetuned model with 7~11k images.

• Better prompt comprehension. Trained with precise and logical natural language captions from Gemini, rather than tags in random order with high FPR.

• Still maximum creativity. No default style. The dataset is very diverse.

• Finest textures, lighting and art details. The dataset only contains images with the highest quality.

• Zero plastic glossy AI style feelings. Glossy Al styles and images are polluting the world, but not on my watch, I handpicked every single image.

• Calibrated contrast and lighting. Say goodbye to overflowing and oversaturation and CFG hacks (RescaleCFG etc.).

• Standard noise scheduler. Inpainting/img2img friendly.

Why no default style?

What is "default style"? If a model has a default style (bias), it means no matter what you prompted, the model must generate the same things (faces, backgrounds, feelings) that make up the default style.

• Pros: Easy for beginners. You won't have to prompt style and you will always get a stable style.

• Cons: If you are a advanced user, you will notice that you can't overwrite the default style. If you prompt something that does not fit the default style, the model will just ignore it. If you stack more styles, the default style will always overlap/pollute/limit other styles.

"no default style" means no bias. The model is very creative and you can get the style/things you prompted exactly as it should be.

Improvements:

Now the model can accurately generate the style you prompted. No style overlapping/shifting, no AI faces, only better details. See comparison (with base model):

• https://civitai.com/images/84145167 (general styles)

• https://civitai.com/images/84256995 (artist styles)

See more xy plots in cover images.

Why is this "finetuned base model" a LoRA?

• This LoRA is actually a DoRA (from Nvidia), which is more efficient than traditional LoRA.

• Enough for thousands of training images.

• 60MiB LoRA vs 7GiB big fat checkpoint. 99% smaller, I call it efficiency.

Share merges using this model is prohibited. FYI, there are hidden trigger words to print invisible watermark. I coded the watermark and detector myself. I don't want to use it, but I can.

This model only published on Civitiai and TensorArt. If you see "me" and this sentence in other platforms, all those are fake and the platform you are using is a pirate platform.

How to use

This is a finetuned base model but in a LoRA.

Just use the pretrained base model and load this LoRA with strength 1.

Also, if you are loading multiple LoRAs, make sure this LoRA is the first in your LoRA stack. This LoRA uses a new arch called DoRA from Nvidia, more efficient than traditional LoRA. But the patch weight is dynamically calculated based on the currently loaded base model weights (which changes when you loading LoRAs). To minimize the unexpected changes, load this LoRA first.

Versions:

• cknb (trained on ChenkinNoob-XL).

• nbvp10 (trained on NoobAI v-pred v1.0). Note: Calibrated noise scheduler. Balanced contrast and lighting. No overflowing and oversaturation and you don't need CFG hacks (RescaleCFG etc.).

• nbep10 (trained on NoobAI eps v1.0).

• illus01 (trained on Illustrious v0.1).

Merge it to other finetuned models:

This is a LoRA after all, but be aware:

It does not work on AI style polluted 1girl overfitted 50 versions of Nova furry 3D anime WAI or whatever. It can't improve those glossy shiny plastic AI style. Use a pretrained base model if you want to get rid of AI style.

• This is what Craft Lawrence (from spice and wolf) should be, if you've seen the anime: https://civitai.com/images/107381516

• This is what those AI style polluted 1girl overfitted model generated: https://civitai.com/images/107647042

If it destroyed your base model even at low strength (e.g. <0.5). That's your base model's problem. Your base model has already merged this LoRA (and you merged it twice). And the model weights got multiplied and collapsed. Beware of fake base model creators, aka. thieves. Some "creators" never do the training, they only grab other people's models, merge them, wipe all metadata and credits, and sell it as their own base model.

Old versions:

You can find more info in "Update log". Beware that old versions may have very different effects.

Main timeline:

• Now ~: Natural details and textures, stable prompt understanding and more creativity. Not limited to pure 2D anime style anymore.

• illus01 v1.23 / nbep11 0.138 ~: Better anime style with vivid colors.

• illus01 v1.3 / nbep11 0.58 ~: Better anime style.

Update log

(12/26/2025): cknb02 v0.299a

"a" means it has a weak default anime style.

Dataset changed a lot, I think ~50% images are new and all captions are regenerated by Gemini 3.

(10/21/2025): Noobai v-pred v0.280a

Special version, "a" means anime. There is a default 2D anime style. So it should be easier to use if you don't want to prompt style. Also the dataset changed a lot, the effect may quite different than previous version.

(8/31/2025) NoobAI ep10 v0.273

This version is trained from the start on NoobAI eps v1.0. Better and balanced brightness in extreme conditions.

(8/24/2025) NoobAI v-pred v0.271:

Comparing to previous v0.264:

• Better and balanced lighting in extreme condition, less bias.

• High contrast, pure black 0 and white 255 in the same image, even at same place, no overflowing and oversaturation. Now you can have all of them at once.

• (old v0.264 will try to cap the image between 10~250 to avoid overflowing, and still has noticeable bias issue, overall image may be too dark or bright)

(8/17/2025) NoobAI v-pred v0.264:

First version trained on NoobAI v-pred. Better lighting, less overflowing.

(7/28/2025) illus01 v1.198

Mainly comparing to v1.185c:

• End of "c" version. Although "visually striking" is good but it has compatibility issues with other style LoRAs.

• Instead, more textures and details. Cinematic level of lighting. Better compatibility.

• This version changed lots of things, including dataset overhaul, so the effect will be quite different than previous versions.

• For those who want v1.185c crazy effects back. You can find pure and dedicated art styles in this page. If dataset is big enough for a LoRA, I may train one.

(6/21/2025) illus01 v1.185c:

Comparing to v1.165c.

• +100% clearness and sharpness.

• -30% images that are too chaotic (cannot be descripted properly). So you may find that this version can't give you a crazy high contrast level anymore, but should be more stable in normal use cases.

(6/10/2025): illus01 v1.165c

• This is a special version. This is not an improvement of v1.164. "c" stands for "colorful", "creative", sometimes "chaotic".

• The dataset contains images that are very visually striking, but sometimes hard to describe e.g.: Very colorful. High contrast. Complex lighting condition. Objects, complex pattens everywhere.

• So you will get "visually striking", but at cost of "natural". May affect styles that have soft colors, etc. E.g. This version cannot generate "pencil art" texture perfectly like v1.164.

(6/4/2025): illus01 v1.164

• Better prompt understanding. Now each image has 3 natural captions, from different perspective. Danbooru tags are checked by LLM, only important tags are picked out and fused into the natural caption.

• Anti-overexpose. Added a bias to prevent model output reaching #ffffff pure white level. Most of the time #ffffff == overexposed, which lost many details.

(5/19/2025): illus01 v1.152

• Continual to improve lighting and textures and details.

• 5K more images, more training steps, as a result, stronger effect.

(5/9/2025): nbep11 v0.205:

• A quick fix of brightness and color issues in v0.198. Now it should not change brightness and colors so dramatically. v0.198 isn't bad, just creative, but too creative.

(5/7/2025): nbep11 v0.198:

• Added more dark images. Less deformed body, background in dark environment.

• Removed color and contrast enhancement. Because it's not needed anymore. Use Contrast Controller instead.

(4/25/2025): nbep11 v0.172.

• Same new things in illus01 v1.93 ~ v1.121.

• Better natural texture, background, lighting. Weaker character effects for better style compatibility.

• Better color accuracy and stability.

(4/17/2025): illus01 v1.121.

• Rolled back to illustrious v0.1. illustrious v1.0 and newer versions were trained with AI images deliberately (maybe 30% of its dataset). Which is not ideal for LoRA training. I didn't notice until I read its paper.

• Lower character style effect. Back to v1.23 level. Characters will have less details from this LoRA, but should have better compatibility. This is a trade-off.

• Other things just same as below (v1.113).

(4/10/2025): illus11 v1.113

Trained on Illustrious v1.1.

• More data for texture, lighting and depth of field effect. Better background structural stability. Less deformed background, like deformed rooms, buildings.

• Full natural language captions from LLM.

(3/30/2025): illus01 v1.93.

• v1.72 was trained too hard. So I reduced it overall strength. Should have better compatibility.

(3/22/2025): nbep11 v0.160.

• Same stuffs in illus v1.72.

(3/15/2025): illus01 v1.72

• Same new texture and lighting dataset as mentioned in ani40z v0.4 below. More natural lighting and natural textures.

• Added a small ~100 images dataset for hand enhancement, focusing on hand(s) with different tasks, like holding a glass or cup or something.

• Removed all "simple background" images from dataset. -200 images.

• Switched training tool from kohya to onetrainer. Changed LoRA architecture to DoRA.

(3/4/2025) ani40z v0.4

• Trained on Animagine XL 4.0 ani40zero.

• Added ~1k dataset focusing on natural dynamic lighting and real world texture.

• More natural lighting and natural textures.

ani04 v0.1

• Init version for Animagine XL 4.0. Mainly to fix Animagine 4.0 brightness issues. Better and higher contrast.

illus01 v1.23

nbep11 v0.138

• Added some furry/non-human/other images to balance the dataset.

nbep11 v0.129

• bad version, effect is too weak, rank is too small, just ignore it

nbep11 v0.114

• Implemented "Full range colors". It will automatically balance the things towards "normal and good looking". Think of this as the "one-click photo auto enhance" button in most of photo editing tools.

• One downside of this optimization: It prevents high bias. You can't generate an image with 95% of the image is dark, and 5% bright. It will be balanced to 50/50%

• Added a little bit realistic data. More vivid details, lighting, less flat colors.

illus01 v1.7

nbep11 v0.96

• More training images.

• Then finetuned again on a small "wallpaper" dataset (Real game wallpapers, the highest quality I could find. ~100 images). More improvements in details (noticeable in skin, hair) and contrast.

nbep11 v0.58

• More images. Change the training parameters as close as to NoobAI base model.

illus01 v1.3

nbep11 v0.30

• More images.

nbep11 v0.11: Trained on NoobAI epsilon pred v1.1.

• Improved dataset tags. Improved LoRA structure and weight distribution. Should be more stable and have less impact on image composition.

illus01 v1.1

• Trained on illustriousXL v0.1.

nbep10 v0.10

• Trained on NoobAI epsilon pred v1.0.