Raehoshi Illust XL
an enhanced iteration built upon the Illustrious XL model. It aims to elevate the visual style by addressing some of the limitations in the original, such as oversaturation and artifact noise. While these issues are not entirely eliminated, noticeable improvements have been made. The goal is to deliver a more polished, balanced output while staying true to the strengths of the base model.
Why Early Access?
Early access helps keep the project going. I don’t have my own GPU, so all training is done through rented cloud GPUs and that gets pretty expensive. By getting early access, you’re directly supporting the development of my models and helping me keep improving them. If you'd like to support me further, you can also buy me a coffee on Ko-fi! Every bit of help means a lot and keeps the future updates coming.
Recommended setting
Positive prompt :
masterpiece, best quality, very aesthetic, absurdresNegative prompt :
bad quality, worst quality, jpeg artifacts, sketch, bad anatomy, signature, watermarkSteps : 25+
CFG : 5-7
Sampler : euler a or dpm++2m karras (euler for vpred)
Standard resolution :
832 x 1216, 1216 x 832, 1152 x 896, 896 x 1152, 1344 x 768, 768 x 1344, 1024 x 1024High resolution :
1024 x 1536, 896 x 1536, 1536 x 1024, 1536 x 896Hires.fix Setting:
Upscaler : 4x Foolhardy Remacri
Hires step : 10-15
Denoise : 0.1-0.3
Special Tags
Quality Tags:
masterpiece
best quality
good quality
average quality
bad quality
worst quality
Rating Tags:
safe
sensitive
nsfw
nsfw, explicit
Aesthetic Tags:
very aesthetic
aesthetic
displeasing
very displeasing
Training Details
The model was developed using a two-stage fine-tuning process. In Stage 1, new series and characters were introduced into the model. Stage 2 focused on fixing issues and enhancing the overall style for improved output.
Stage 1
Dataset : v1-31k, v2-37k, v3-34k, v4-60k, v5_v5.1-18k, v6-15k, v7-39k, v8-41k, v9-30k, v10-30k with multi resolution
Hardware : 2x A100 80gb, v3, v4, v5, v5.1-2x H100 80gb, v7,v8, v9, v10-RTX PRO 6000
Batch size : 32
Gradient accumulation steps : 2
Learning rate : 6e-6
Text encoder : 3e-6
Epoch : 15
Stage 2
Dataset : v1-2.5k, v2 and v3-2.3k, v4-2.5k, v5-2k, v5.1-1.8k, v6-1.5k, v7-1.7k, v7.1,v8-4.1k, v9-1.9k, v10-2.4k
Hardware : 1x A100 80gb, v7_v7.1,v8, v9, v10-RTX PRO 6000
Batch size : 48
Gradient accumulation steps : 1
Learning rate : 3e-6, v5.1-2.5e-6
Text encoder : disable
Epoch : 15
List of New Series/Characters Trained:
Zenless Zone Zero
Wuthering Waves
Honkai: Star Rail
Genshin Impact
Arknights: Endfield
Umamusume
Azur Lane
Arknights
Fate/GO
Dandadan
Make heroine ga oo sugiru
Kusuriya no Hotorigoto
Hololive from justice and dev is
Indie Vtuber Dooby, Yuuki Sakuna, Nimi Nightmare, and S***
100 girlfriends who really love you
Haite kudasai takamine-san
Alina clover
Nikke: bready and little mermaid
Kpop Demon Hunters
Full character list are available article here
For character trait details prompts, please refer to the Danbooru site for accurate tags and references.
License
Special thanks to Joe for supporting my works
Special thanks to Juno for supporting my works and help me with early tester
Description
FAQ
Comments (8)
An amazing model with a very good prompt adherence!, the only lets say "bug", i'm getting is that sometimes the subjects gets some neon colors or blu/pink colors in the skin, I fixed it with negatives but I thing is worth mentioning tho
Thank you for mentioning it
I think I found the issue, it was with my default negative prompt. However, by moving 'lowres' to the front of the negative prompt seems to fix the hue problem. I'm still not sure why this has such an effect.
@Raelina interesting, glad you have an idea of what i'm talking about, btw I have found another "bug", happened like twice but when asking a naked subject, the skin tends to go all grey like if it was using a tight suit, anyway, I will try your solution because the quality of the results with this one is insanely good!, feels like 4k resolution always lol
I was playing with this model but I notice it will try to do some angel like glow on the background of the image or elaborated designs and that is even when I specify the background.
Not sure if "intricate details, elaborated designs" prompts are the cause, something else. Either way, wanted to point out that, also I notice some series lose their "artstyle" when you don't have an artist, compared to original IllustriousXL
Either way, still doing testing and having fun, might post something when I get a result I really like. Keep it up, updating the model with the latest characters is always a good move!
Thank you for your feedback
This is a model I really enjoy using because I love the style, and in many cases, the anatomy is even better than with the pony model.
However, I've noticed some issues (mostly in the NSFW content) that make it weaker compared to the Pony model. I've generated over 500+ images using this model across various concepts, and here are a few weak points in NSFW concepts:
- Fellatio POV
- Standing sex from behind / arms pulled behind
- Inconsistent feet
Additionally, I've found that inpainting doesn’t work as well as it does in Pony models. It feels like this model doesn't understand what I want. For example, when I try to correct a missing nail by drawing a colorful one and masking the area, the pony model typically picks it up easily, even with a strength setting of 0.35 or a bit higher. But with this model, it keeps skipping the nail and producing inaccurate results. I’m no expert, but it seems like Illustrious (and its finetunes) don't recognize parts of the image as well as Pony models do, which makes fixing details like limbs, nails, or pupils more challenging for me.
Thank you for the detailed review and feedback. This helps me a lot