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
Update Character list knowledge up to Oct 2025:
Azur Lane
Arknights
Fate/GO
Zenless Zone Zero
Genshin Impact
Stella Sora
Nikke: Little Mermaid
Note:
3 games (Azur Lane, Arknights, and Fate/GO) have been added. I’m not sure about the list of character release date since I don’t play those games. Most of the list came from wiki/gemini, so there might be some missing or newer characters, or ones with very few fanarts available. Let me know if you notice any missing characters.
FAQ
Comments (14)
Will you add more characters from earlier gacha games like Honkai Impact 3rd or Punishing: Gray Raven, etc? (shortly, update the database)
Sure, I will try to include them in the next major update
pretty nice, waiting for a downloadable v-pred version
I have tried training my style lora on various illustrious models, but this one is simply head and shoulders above the rest. The only problem is that no controlnets work (tried illustrious XL canny/depth) is there any workaround?
I dont know, I dont really use controlnet much. You can try using noob controlnet, but Im not sure if it will work or not
It seems like an interesting model to me.
Can you make the V-Pred downloadable?
Do you have a list of characters trained in this model?
I actually have the full list in my dataset, but I’m too lazy to extract only the character names right now since there are just too many. I will try to make it later when I have some free time.
Just a heads-up: prompting with only the character name won’t really work. You still need to include their key features like hair color, hairstyle, eye color, accessories, etc. For now, you can check Danbooru for the characters details to help with prompting.
@Raelina Yes, I already use Danbooru, mainly to know which dogs have been trained and which haven't, saving me time. By the way, this model is very good, congratulations.
@Qweser To easily check whether a character can be generated or not, you can look at how many images they have at Danbooru. If a character has fewer than 100 images, it will usually be hard to generate or might only capture some features. I recommend choosing characters with at least 150–200 or more images, as they’re much easier to generate. Thank you
@Raelina I'm also curious to know that information. I've been using your model as a standard since I started using Illustrious. Thank you in advance for your good work with this Checkpoint.
I tested V8.0 and looks like it has some weird... features.
For example, it likes to go "wide shot" mode even when I did not asked for it. Somehow tag "window" triggers this kind of behaviour.
But, as for me, the main problem lies with internal Illustrious styles. Many of these in this checkpoint are disfigured or have very bad anatomy problems (too long bodies), which can not be corrected by negatives. For example, "gusu" and "sadamoto yoshiyuki". Yeah, I know the last one has similar issue on other checkopoints, but for this particular checkpoint it's just plain awful,
The other problems include eyes. Looks like tags "night, dark" just makes eyes worse (more empty).
Also this checkpoint has a very strong bias to "no pupils" tag and to "fine fabric emphasis".
Thank you for the detailed review. This issue might be happening because v8 was trained using a different method than usual, and its also closer to the base model since it only have a small finetune after the pretraining. I will try to improve it in the next update. Thank you