v-pred_v2.0
Carefully Curated Danbooru Dataset (up to September 2025)
Base Model: NoobAI-XL (NAI-XL) V-pred-1.0-Version
When necessary, please use Rescale CFG with values up to around 0.7 (especially for NSFW content).
Quality Tags:
very aesthetic > masterpiece > best quality > good quality > normal quality > worst quality
If you feel like it, you can support me on Ko-fi. Thank you.
Online Generator:https://www.seaart.ai/ja/models/detail/a106ce42854f7b7f65523ad10e03caa8
v-pred_v1.1
Dataset:50k AI-generated images from Aibooru and Pixiv + 10k images from Danbooru
Prompts
people, characters, copyright, style, general tags, rating, masterpiece, best quality, good quality, newest
Negative Prompts
lowres, worst quality, bad quality, bad anatomy, sketch, jpeg artifacts, signature, watermark, old, oldest
Base model:NoobAI-XL (NAI-XL) V-pred-1.0-Version + Illustrious XL 1.0
v-pred_v1.0
Base model:NoobAI-XL (NAI-XL) V-pred-1.0-Version
Use v-prediction and Zero Terminal SNR.
For information on v-prediction and how to use it in ComfyUI, refer to the NOOBAI XL Quick Guide by L_A_X.
As of November 20, checkpoints supporting v-prediction are not available for use in stable-diffusion-webui.
Outside of ComfyUI, you can use the webui's dev branch (likely requiring a .yaml file for each checkpoint) or Forge and reForge.
Additionally, make sure to select "Zero Terminal SNR" under Settings → Sampler parameters → Noise schedule for sampling.
v3.1
Base model:NoobAI-XL (NAI-XL) Epsilon-pred 1.1-Version
v-pred_v0.1
Base model:NoobAI-XL (NAI-XL) V-pred-0.6-Version
v3.0
Base model:NoobAI-XL (NAI-XL) Epsilon-pred 1.0-Version
v2.1
Base model:NoobAI-XL(noobaiXLNAIXL_epsilonPred075 x 0.7 + noobaiXLNAIXL_epsilonPred05Version x 0.3)
v2.0
Base model:NoobAI-XL (NAI-XL) Epsilon-pred 0.5-Version
v1.0
Base model:Illustrious-XL-v0.1
After training, this model merged Lucereon into the U-net.
(0,0.3,0.25,0.2,0.15,0.1,0.15,0.2,0.25,0.3,0.3,0,0.05,0.15,0.25,0.3,0.25,0.15,0.05,0)
License
Description
FAQ
Comments (38)
Does Vpred11 still require Danbooru-style prompting, or can we freely use any words outside of Danbooru now?
There is nothing special about v-pred that would've allowed that, NoobAI is all about booru tags
Love you. Version 1.1 incorporates the high-resolution features of Illustrious 1.0, while retaining the original color depth of NoobAI. The prompt words exhibit high compliance, and there's a significant improvement in spatial and limb coordination. The biggest surprise for me is that the quality tags now also include male weighting.
do you have plan for eps version update?
I don't know if it's user error or not, but I'm having all kinds of anatomy issues with the 1.1 version.
v1.1 is so much more creative and responsive! But the anatomy is a little bit less stable, and the color is a little washed out compared to previous version
I genuinely cannot control the anatomy with 1.1 lol
are you use a comfly ui? I found that forge has strange anatomy problems
V1 was my favorite Illustrious model and this one is going to steal it's place.
I'm having one issue though, enabling hires fix or Addetailer completely butchers the picture.
Can anyone share their settings for hires fix or addetailer? I'm on Forge btw and I believe this is 100% an error on my part. I just wanna make sure I use the correct settings
i had something like this happening when the resolution got too high, try lowering the upscaled amount, maybe that helps
@Bumsky no dice, I'm still not getting good results.
I've not tried out 1.1 yet, but in my experience the vpred models need much lower noise levels when doing hi-res fix, otherwise you end up with distortions or complete scene changes. Perhaps Illustrious v1.0 is making that worse.
@Kaligo731 I've been wondering if it's an Illustrious v1 thing as well, I'd need to try out other models using v1
FIGURED IT OUT! I was using Karras as my scheduler type, switching to Automatic IMMEDIATELY fixed it for me.
@Eureka_vi Oh wow! I'll give it a try too, thanks for the tip.
The GOAT is back!!! I'm clearing out my whole weekend fellas!!
Idk what I'm doing wrong, but the model tends to censor every time. No matter what kind of censor tag I put into negative prompt. It's mostly looks like mosaic one, but nothing help to get rid of it.
put uncensored in main prompt
Will there be EPS version update? V-pred is great but it tends to emphasize the environment instead of human figure, which does not always fit the expectation...
It seems like models that use illustrious 1.0 are a lot more likely to completely mess up anatomy, that being said, it's still a great model, and I tend to prefer V-pred, so this is absolutely my go-to model now.
You have to work harder to get perfection, but, if you dedicate the time, then It's FAR better than an older illustrious version model. Great work!
OB is my favorite checkpoint, so I created a comparison chart between VPRED1 and 1.1,
V 1.1 is strange - sometimes it performs excellently, while other times it's terrible. But calling it unstable feels like an overstatement. I don't know how to properly assess it. The only thing certain is that it's not a comprehensive upgrade over version 1.0.
这模型出的跟tm王炸一样
This is a great model! If you don't mind, I have a suggestion for the next version: Add training dataset images from Furbooru, Derpibooru, Twitter, and DeviantArt.
better than ntr mix, mostly due to better lighting
If this model gets updated again, I could ascend to heaven right on the spot... ( ˙▿˙ )
Amazing model! Although this model is not as good as WAI in anatomy and fingers, it possesses an unparalleled advantage: the generated images feature natural, hand-drawn-like lines, making them appear more like paintings rather than AI creations. This effect is truly fantastic, so I immediately switched from WAI. I have tried dozens of models and have yet to find any that can achieve this effect.
I very much agree.
这模型太牛逼了，除了画手容易崩其他都挺无敌的，赶快更新吧求求了
最爱的模型没有之一。很好保留了原始人物特征，适当选择艺术家和调整画风可以消除ai感且风格多变，皮肤光影质感顶级
yo are u gonna do any updates, every other finetune sucks
best one for sure
Fingers and toes are easily to get messy comparing to wai noob or one obsession noob.
最高のモデル！よければアップデートをお願いします！
best model ever.
A little hand failure, but still worth using.
I will be very happy if updated this
EXTREMELY UNDERRATED
Any plans for an update?
Would love to see more of this model!
Very good model, but fingers tend to get bad
The extremely underestimated model generates a pseudo 2.5D effect that is better than other models
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.