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, v11-29k
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, v11-3k
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 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
Improved overall quality
Based on NoobAI 1.1
Updated dataset with new series "makeine"
Special tags "dramatic lighting" (same as v2)
Include cyberfix
FAQ
Comments (20)
Hi, great work! ¿Could you please enable online generation to test it? Thanks!
I like version 2.1 more than 3.0
Thank you for the review and making a comparison. V3 is built on a completely different base than V2.1. While V2.1 is still aligned with Illustrious, V3 is not, so the outputs will naturally differ.
what's the advantage of this model over animagine/rae-animagine? ive been using it (never failed me) for a long time and wondering what's out there
This model was trained with a larger dataset and recognizes many characters. If you enjoy using artist tags or combining them, this model offers more advantages. However, if you prefer an animated style, stick to using Animagine or Rae Diffusion instead, as they are also easier to use
@Raelina I don't use artist tags because I only use my own art, so mostly LoRAs. I'm sticking to animagine then while waiting for the next version :P
There's a problem with the license for this. It doesn't include NoobAI-XL's addendum to the Illustrious license.
Please consider creating a fine-tune of Illustrious-XL independent of noob, the license for NoobAI-XL is extremely limiting to the point where posting images without prompts is a violation of the license.
what?? Are you serious??
@Raelina From https://civitai.com/models/833294/noobai-xl-nai-xl
"III. Open Source Community
To foster a thriving open-source community,users MUST comply with the following requirements
Open source derivative models, merged models, LoRAs, and products based on the above models.
Share work details such as synthesis formulas, prompts, and workflows.
Follow the fair-ai-public-license to ensure derivative works remain open source."
"MUST comply with the following requirements... ...share work details such as synthesis formulas, prompts, and workflows." Users of NoobAI-XL and derivatives are compelled by this license to share "...prompts, workflows". It is a requirement.
noobai is also seemingly released under TWO licenses because there is the license spelled out on the CivitAI description, and then there is the addendum license here: https://civitai.com/models/license/1140829
While the restrictions laid out between the two licenses are similar, they're not the same. It's confusing but the limitations placed are still extreme no matter which way you look at it.
The FAIPL license is all about open-sourcing the model. If they impose restrictions of this nature, I will discontinue my support for it. This update will be my last using NoobAI as the base. I'm going back to illustrious base
@Raelina I think there are some very strong innovations in noob so it's a shame. I'm sorry I have to be the one to bring you this bad news. While I have no problem with sharing my own prompts and go out of my way to do so, I agree it's a mistake to put restrictions like this on outputs.
Thank you for continuing to put out excellent fine tunes.
@Raelina I mean even if so, then how are they supposed to enforce this? Chase after anonymous 200,000 people on the internet? I'd say you should keep using noob, it's much more likely they'll change the license instead of going after users. That would ruin their reputation
@DevilLady I agree it's basically impossible to enforce but if they're going to change it, why use this restrictive license in the first place? It makes no sense.
@ArtificialSweetener_ Let's hope more people will post about this and they may change it
Any tips on how to get a more darker, nighttime setting? the images always are very bright/saturated even with prompts like "dark background, night," etc."
It's the best model I've ever used. Very excited about possible new versions with more characters introduced.