Details
Related Works
V3 (Il1.0)
ILv1.0
GR-Illustrious Semi Realistic NIJI

GR-Illustrious Semi Realistic NIJI

Exclusive
Original
4.7M
9.3K
0
#Realism
#Model Anime
#negy
#✨checkpoint
#Base Model
#Furry
#Illustrious
...View More

Semi Realistic NIJI model is a model that excels at semi-realism and aesthetic NIJI features as you can understand from name. Model is trained to serve for people who doesn't like standart anime generations yet want to have features of anime in their realistic generations. It performs quite well with lighting prompts so it's recommended to use them in prompts. Certain anime style LoRAs are not recommended to use since model itself has a style. Other than that, model can perform well with LoRAs. Dataset can be older than latest datasets since it's hard to maintain style of model while adding new datasets to model, yet model still can generate more characters than regular Illustrious models and it's Illustrious XL 1.0 based so it supports higher resolutions like 1536px while standart Illustrious XL models support 1024px based resolutions.

Please use Normal Mode while generating, Quality Mode increase the resolution and deforms the image. Correct resolution options are already selected. Also Normal Mode is cheaper. So you get better images for cheaper price. Please select it before using the model.

View Translation
GrayMan
5.1K
57.0K
Chat with the model
Notice
2025-03-25
Publish Model
2025-11-25
Add Template Version
2026-02-25
Update Model Info
Model Details
Type
Checkpoint
Publish Time
2025-11-25
Base Model
Illustrious
Trigger Words
Copy
Model Parameters
View More
Training Parameters
Epochs：1
Steps：0
Version Introduction

GR-Illustrious Semi Realistic NIJI (ILv1.0)

This is a Illustrious XL 1.0 based semi realistic NIJI style model. Model already has unique style, it's not recommended to use with style LoRAs. Model works great with LoRAs and has knowledge of IL1.0 base model for characters. In this update i changed style a little bit and extended dataset, also improved quality and accuracy. If you wonder why i changed style, i was looking at user generations and noting what type style they generate on this model, so i edit style a little bit closer to what users generally wants to create. If you don't like the changes, you can always keep using the older version.

If you liked model, you can favourite it and give it a 5 star rating. It keeps me motivated. Thanks!

Recommended Resolution

This is the most important part so please read carefully. This model is based on IL1.0 base model and resolution choices are higher than standart models. If you use 768x1152 like standart resolutions offered by SeaArt, quality will significantly drop and you won't be able to see actual quality of model. Please use these resolutions for generation.

1:2 - 2:1 768x1536 - 1536x768

9:16 - 16:9 864x1536 - 1536x864

2:3 - 3:2 1024x1536 - 1536x1024

3:4 - 4:3 1152x1536 - 1536x1152

4:5 - 5:4 1152x1440 - 1440x1152

1:1 1536x1536

Recommended Positive: masterpiece, best quality, amazing quality, hyper-detailed, newest, very aesthetic

Recommended Negative: lowres, bad quality, worst quality, bad hands, bad , sketch, jpeg artifacts, ugly, poorly drawn, censor, blurry, watermark, artistic failure, artistic error, bad proportions, bad perspective, displeasing, very displeasing, oldest, child, childish

Note: This model has furry support unlike the standart Illustrious Models. Use `1female, {animal type} anthro` in prompt to create furry arts. It can also work with spesific Danbooru tags like `lizard girl, dragon girl` etc.

Recommended Settings

You can use CFG value between 3~7, CFG 5 is Recommended for standart usage.

25~40 Steps, 30 is sweetpoint for model.

Use Euler A Sampler with model.

Model has baked-in vae, no need to use any vae.

ClipSkip must be 2.

Use Creative Upscale to adjust small details and refine image.

Thank you for reading!