Model Introduction
This image generation model, based on Laxhar/noobai-XL_v1.0, leverages full Danbooru and e621 datasets with native tags and natural language captioning.
Implemented as a v-prediction model (distinct from eps-prediction), it requires specific parameter configurations - detailed in following sections.
Special thanks to my teammate euge for the coding work, and we're grateful for the technical support from many helpful community members.
⚠️ IMPORTANT NOTICE ⚠️
THIS MODEL WORKS DIFFERENT FROM EPS MODELS!
PLEASE READ THE GUIDE CAREFULLY!
Model Details
Developed by: Laxhar Lab
Model Type: Diffusion-based text-to-image generative model
Fine-tuned from: Laxhar/noobai-XL_v1.0
Sponsored by from:
Collaborative testing:
How to Use the Model.
Guidebook for NoobAI XL:
ENG:
https://civarchive.com/articles/8962
CHS:
https://fcnk27d6mpa5.feishu.cn/wiki/S8Z4wy7fSiePNRksiBXcyrUenOh
Recommended LoRa List for NoobAI XL:
https://fcnk27d6mpa5.feishu.cn/wiki/IBVGwvVGViazLYkMgVEcvbklnge
Method I: reForge
(If you haven't installed reForge) Install reForge by following the instructions in the repository;
Launch WebUI and use the model as usual!
Method II: ComfyUI
SAMLPLE with NODES
Method III: WebUI
Note that dev branch is not stable and may contain bugs.
1. (If you haven't installed WebUI) Install WebUI by following the instructions in the repository. For simp
2.Switch to dev branch:
git switch dev
3. Pull latest updates:
git pull
4. Launch WebUI and use the model as usual!
Method IV: Diffusers
import torch
from diffusers import StableDiffusionXLPipeline
from diffusers import EulerDiscreteScheduler
ckpt_path = "/path/to/model.safetensors"
pipe = StableDiffusionXLPipeline.from_single_file(
ckpt_path,
use_safetensors=True,
torch_dtype=torch.float16,
)
scheduler_args = {"prediction_type": "v_prediction", "rescale_betas_zero_snr": True}
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, **scheduler_args)
pipe.enable_xformers_memory_efficient_attention()
pipe = pipe.to("cuda")
prompt = """masterpiece, best quality,artist:john_kafka,artist:nixeu,artist:quasarcake, chromatic aberration, film grain, horror \(theme\), limited palette, x-shaped pupils, high contrast, color contrast, cold colors, arlecchino \(genshin impact\), black theme, gritty, graphite \(medium\)"""
negative_prompt = "nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro"
image = pipe(
prompt=prompt,
negative_prompt=negative_prompt,
width=832,
height=1216,
num_inference_steps=28,
guidance_scale=5,
generator=torch.Generator().manual_seed(42),
).images[0]
image.save("output.png")
Note: Please make sure Git is installed and environment is properly configured on your machine.
Recommended Settings
Parameters
CFG: 4 ~ 5
Steps: 28 ~ 35
Sampling Method: Euler (⚠️ Other samplers will not work properly)
Resolution: Total area around 1024x1024. Best to choose from: 768x1344, 832x1216, 896x1152, 1024x1024, 1152x896, 1216x832, 1344x768
Prompts
Prompt Prefix:
masterpiece, best quality, newest, absurdres, highres, safe,
Negative Prompt:
nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro
Usage Guidelines
Caption
<1girl/1boy/1other/...>, <character>, <series>, <artists>, <special tags>, <general tags>, <other tags>
Quality Tags
For quality tags, we evaluated image popularity through the following process:
Data normalization based on various sources and ratings.
Application of time-based decay coefficients according to date recency.
Ranking of images within the entire dataset based on this processing.
Our ultimate goal is to ensure that quality tags effectively track user preferences in recent years.
Percentile RangeQuality Tags> 95thmasterpiece> 85th, <= 95thbest quality> 60th, <= 85thgood quality> 30th, <= 60thnormal quality<= 30thworst quality
Aesthetic Tags
TagDescriptionvery awaTop 5% of images in terms of aesthetic score by waifu-scorerworst aestheticAll the bottom 5% of images in terms of aesthetic score by waifu-scorer and aesthetic-shadow-v2......
Date Tags
There are two types of date tags: year tags and period tags. For year tags, use year xxxx format, i.e., year 2021. For period tags, please refer to the following table:
Year RangePeriod tag2005-2010old2011-2014early2014-2017mid2018-2020recent2021-2024newest
Dataset
The latest Danbooru images up to the training date (approximately before 2024-10-23)
E621 images e621-2024-webp-4Mpixel dataset on Hugging Face
Communication
QQ Groups:
427280545
677964513
852429527
914818692
635772191
870086562
Discord: Laxhar Dream Lab SDXL NOOB
How to train a LoRA on v-pred SDXL model
A tutorial is intended for LoRA trainers based on sd-scripts.
article link: https://civarchive.com/articles/8723
Utility Tool
Laxhar Lab is training a dedicated ControlNet model for NoobXL, and the models are being released progressively. So far, the normal, depth, and canny have been released.
Model link: https://civarchive.com/models/929685
Model License
This model's license inherits from https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0 fair-ai-public-license-1.0-sd and adds the following terms. Any use of this model and its variants is bound by this license.
I. Usage Restrictions
Prohibited use for harmful, malicious, or illegal activities, including but not limited to harassment, threats, and spreading misinformation.
Prohibited generation of unethical or offensive content.
Prohibited violation of laws and regulations in the user's jurisdiction.
II. Commercial Prohibition
We prohibit any form of commercialization, including but not limited to monetization or commercial use of the model, derivative models, or model-generated products.
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.
IV. Disclaimer
Generated models may produce unexpected or harmful outputs. Users must assume all risks and potential consequences of usage.
Participants and Contributors
Participants
L_A_X: Civitai | Liblib.art | Huggingface
li_li: Civitai | Huggingface
nebulae: Civitai | Huggingface
Chenkin: Civitai | Huggingface
Euge: Civitai | Huggingface | Github
Contributors
Narugo1992: Thanks to narugo1992 and the deepghs team for open-sourcing various training sets, image processing tools, and models.
Onommai: Thanks to OnommAI for open-sourcing a powerful base model.
V-Prediction: Thanks to the following individuals for their detailed instructions and experiments.
adsfssdf
madmanfourohfour
Community: aria1th261, neggles, sdtana, chewing, irldoggo, reoe, kblueleaf, Yidhar, ageless, 白玲可, Creeper, KaerMorh, 吟游诗人, SeASnAkE, zwh20081, Wenaka~喵, 稀里哗啦, 幸运二副, 昨日の約, 445, EBIX, Sopp, Y_X, Minthybasis, Rakosz, 孤辰NULL, 汤人烂, 沅月弯刀,David, 年糕特工队，
Description
This week's update features the 60% version of the V-prediction model. Over the past week, Laxher Lab has made new progress in many areas, including base model and ControlNet training. Compared to the V-Pred-0.5-Version, this version has many improvements in usage, especially in color saturation. A brief summary of this week's updates is as follows:
1. Improved color saturation generation in previous versions, now it is less likely to overexpose when generating colors like red;
2. Adapted to more sampling methods besides Euler, including Euler a, and other sampling methods will be supported in subsequent versions;
3. Released three new NoobAI XL dedicated ControlNet models: openpose, softedge, and lineart;
4. Supported NoobAI XL's V-prediction version in the main branches of forge and reforge, now you can use the model with one-click automatic background recognition.
5. We have created and organized some introductory guides for NoobAI XL to facilitate new users, which are published at https://fcnk27d6mpa5.feishu.cn/wiki/S8Z4wy7fSiePNRksiBXcyrUenOh. We will also release its English version on civitai soon, and the model's main introduction page will be reformatted in the near future to make it easier to read and get started.
We hope enthusiasts can enjoy the new V-Pred-0.6-Version. Your passion is the driving force behind our updates.
FAQ
Comments (93)
As far as I understand NoobAI-XL is based on Illustrious-XL, so could you please set the base model to display "Illustrious" instead of "SDXL 1.0"? Otherwise great model, thank you for your work! :)
0.6V-pred版本肢体十分糟糕,我退回了0.5
This seems like an innovative model, but what is the goal of this?
Seems like V-pred 0.6 is somewhat a bit downgrade than V-pred 0.5. The obvious drawback is the quality of eyes. V-pred 0.6 tends to be blurry and undetailed compared to V-pred 0.5. Besides, the over saturation issue still exists and seems even worse... Despite these, it's a great step that they've released official controlnet model for noobai xl. Hope tile model will get released soon. Good job!
Didn't notice that, thanks for sharing!
Can you provide some examples? Thank you very much
If I use Noob as a base model to train style lora, what trigger words should I use? Since this style is already included in the model, should I use "artist: XXX" or "XXX"?
@Flange Hi, for any style LoRAs it's better not to use any trigger tag. But it's better not forget about turning on shuffle captions parameter. But if you really need a style tag, just use "XXX", not "by XXX" or "artist: XXX".
@TroubleDarkness Thank you very much! I will try it.
@TroubleDarkness Just because I noticed that the recommended style tag in Noob's guidebook is "artist:XXX", so i have this question about how to deal with the tag during training.
@Flange Actually, info about proper artist tags is extremely unclear, older version says just "name", guidebook says "artist: name", anyway. you can just check the prompts with both "artist: name" and "name". constructions. You'll find they're working pretty much the same way. IMHO, single artist tag without "artist:" works better. So, there is 100% chance that it was trained just like "name" tag.
@TroubleDarkness I think LAX was/is a NAIV3 user so he probably just copies the mixes from there which uses the Artist:XXX syntax. It's not necessary in Noob since it's trained with just name but from my tests, it doesn't really matter, Noob or Illus picks it up just the same.
Is NoobAI-XL based on Illustrious-XL or did you create this model entirely without Illustrious-XL? Because you just deleted the reference from CIvitAI, that NoobAI-XL is "based on training from Illustrious-xl". And I'm not talking about any LoRA incompatibility caused by using V-Pred instead of Epsilon-Pred, which in my opinion has nothing to do with the question. I like NoobAI-XL (so no need to downvote me again) and simply want know if it makes sense that NoobAI-XL gets its own base model tag on CivitAI to easily find fine tunes of this model.
it would make absolute sense for it to get its own tag, but that's up to civit staff
I guess no answer is also an answer... NoobAI-XL is based on Illustrious-XL and therefore should credit it.
Fire
Why is it prompted like this ?
\(azur lane\)
im confused by this, i cant find any explanation or guide to this prompting method
The back slashes "\" are used to escape a character. This is because the parenthesis character is used to trigger something else, and the backslash is to indicate that you want it only as text and not to trigger other actions. As for why you need parenthesis, it's likely the tag or character is ambiguous on its own and needs the text inside the parenthesis to differentiate it.
As you may know, anything in brackets in WebUI will be used with increased emphasis. For example (1girl) is something like ~(1girl:1.11), adding a backslash to the backet symbol just breaks the syntax not to trigger emphasis on tag. Originally, characters and other tags are trained without backslhashes, but nature of emphasis adjustments in the interface obliges you to use backslashes.
okay I get it, I think. I just saw it being used for the series name frequently which confused me.
@shubasaur877 It's just because character names are frequently used in pair with series name in Danbooru tagging system. Danbooru / Gelbooru / rule34 are imageboards for different themed media.
shimakaze (kancolle)
asuna (blue archieve)
etc.
Those franchise suffixes are used not to confuse anyone, because, obviously, there is a lot of characters with the same name or even name and surname at once.
Also, if you're new, don't use tags with underscores, like "asuna_\(blue_archieve\)". Instead of underscores use space while prompting for anything. So, this tag would look like "asuna \(blue archieve\)". There is few exceptions like score tags from a different model or specific artist nicknames, for example.
You may want to install booru tag autocomplete / autocompletion extension for WebUI. It can be found in extensions -> available tab or manuanny downloaded from a github page, using install from URL tab, if you're using A1111, Forge or ReForged WebUIs. You're writing "larg" in the prompt box and you're getting examples of existent tags, like "large breasts", "large bow", "large hat" etc. Extremely helpful for anyone, who's working on Danbooru/E621-based models which NoobAI is.
@TroubleDarkness thank you very much! i understand it completely now. that autocomplete feature seems very interesting however i am using swarm/comfyui, ill look into it, thanks again
@shubasaur877 I believe, SwarmUI has it's own built-in autocompletion feature. Or at least, there is a working extension for it. Not sure about UncomfyUI, nerver was a fan of this quite user-friendly interface.
@TroubleDarkness Thanks i found how to get it working, so much better now and this model is really great :D
@shubasaur877 Good to know.
飞书文档中第三部分的 danbooru角色 的链接错误的指向了和 danbooru艺术家 相同的链接
Hi，各位，我们创建了中文用户手册。手册将提供详尽可靠的模型介绍和使用指南，点击这里查看！
您好，关于文档中提示词规范化那里，我其实一直有个疑问。就是当我使用“颜色+danburoo存在的提示词”的组合的时候，效果是否还能那么好？因为带了颜色的前缀，这个提示词本身就不是一个danburoo存在的提示词了。
另外，指南里只推荐了采样器，是否有推荐的调度器呢？
cool!
@yiliankexi 好问题，一般都是可以的。
模型简介里的“NOOBAI XL 快速指南”文档中提到，艺术家标签需要加artist:前缀，但你的文档里又说艺术家标签不需要添加任何前缀后缀或修饰，以哪个为准？？？
@1519190067765 不需要添加任何前缀。
快速指南应该没有提到需要添加前缀。
@Euge_ 快速指南中“总结”部分的注意事项第三点：提示词使用D站标签，艺术家标签需要添加artist:前缀。我测试的情况是加了artist:效果貌似稍微好一点，是错觉吗
@0xSeiunSky 感谢提醒！以用户手册为准，不需要添加。
其实因为没有直接更改标签文本，因此添加与否不一定影响效果，放心用~
anyone else getting NansException errors in webui? Im on dev branch and using all recommended settings but getting nothing but black screens or jumbled noise
me too
me three
Me too.comfyUI can work.
我也是
all I get is some strange noise, blurred image and oversaturated dots. Using A1111
same, i though my lora do it but nope, run it raw and still only jumbled noise... all other version work totally fine hmm
someone should do embeddings for this model because there is no way to use 1.5 embeddings on comfyui
1.5 embeddings don't work with SDXL
what? you do realize this model is 100% different architecture, right? this is SDXL/Illustrious based, and you don't even need embeddings for 99% of concepts that any normal person would want to use, only if you're trying to use the absolute cutting edge newest anime char/artist that got popular in last 2 months.
Is this model compatible with accelerators like Lightning, Hyper, DMD2, Turbo?
reForged no longer has dev-upstream, they have done some renaming. Which one do you recommend now?
正考慮從ponyxl 轉移至本模型, 但我好奇想問.... 有打算轉換至SD3.5 的基底模型推出新版本嗎?
另外LORA 訓練方面有沒有什麼參考的TAG 或示範數據集可以分享? 至於藝術家方面... LORA訓練時也需要加入相關TAG嗎?
我測試了一個小LORA數據集,我直接套PONY XL的訓練TAG似乎都有很好的效果...好像比PONY XL的效果還好,為什麼可以這麼強
我就觉得大概率不会，不可能一出什么新底膜就马上训练，特别是像flux和sd3.5一类的。sdxl经过几年的各种人的训练才出这样一类实用模型，没有稳定生态的情况下，这种行为是浪费资源的。况且这些模型对显卡要求高，就算有训练出模型，也只不过是少数人的玩具。
lora训练我个人是会使用艺术家名称直接训练的，效果确实会更好
比ponyXL在还原画风上的效果好其实很正常，因为模型本身就保有了对画师图片的训练，lora下的二次训练自然效果会好
据测试群大佬所说，sd3.5为dit架构，相比xl的unet训练所需的算力需要10倍。
How do you use booru tags that already have parenthesis ( ), such as: "champion's_tunic_(zelda)" ?
champion's tunic \(zelda\)
In a word, label normalization is divided into two steps: (i) replacing underscores in labels with spaces and (ii) adding a backslash "\" before brackets.
first you also drop the underscores for spaces. next you escape the parenthesis with \ on the left side of each one.
how to train a lora with this model? i'm trying with sd-scripts yet all i get is blobs
disable all noise related things like noise offset, min snr gamma, multires_noise_discount, and add the follow args --v_parameterization --zero_terminal_snr --scale_v_pred_loss_like_noise_pred --debiased_estimation_loss, the rest of the settings you used to train on other loras like dim etc should be fine.
@Darkwen ok, tks, i'll give it a try
@Darkwen hey it worked, yet i don't know if my configs were wrong, but the Lora did not perform as good as in pony
@Darkwen what about huber schedule = snr?
@Nephilim good to know, what settings are you using? remember that illustrious, and this one specially is very different from pony, pony text encoder is burned to hell.
@Nephilim i dont think huber scheduler matter that much, i didn't change it, you should only use that args and check other settings like the optimizer settings.
@Darkwen dm'ed you
@Nephilim huber is effectively reducting the effect of "crap in, crap out" effect of training (it's an alternative to L2 loss). the paper says it skip bad captionning of subject when it occur in favor of other subjects beter represented. but also it hurt accuracy an convergance.
by affecting it as a schedule, you can effectively dampen the effects of huber loss depending of what you want to train. more detail but more defect vs less details but less defect. depending on your dataset quality this might be relevant, especialy for finetunes which include automated dataset and poisoned pictures.
@Le_Fourbe I actually curate very well my datasets(it's just for LoRA's), so keeping it on must be a bad idea correct?
@Nephilim well it might or might not.
the schedule is by default fairly low to the point you "might" not notice. i didn't run comparision tests myself as i struggle with other parameters too. i keep it on for whatever case one poorly captionned picture creep in.
you might find value in curating just a bit less by using huber loss schedule and find a good good enough result.
After some testing, I have to say this model is BETTER than pony. I thought Pony was pretty much the best, but, this one can do more outlandish concepts where pony will almost always struggle. This will be my go too model from now on, good job guys!
I do find you have to prompt a little 'harder' in this one, than pony, but overall it'll understand what you're asking eventually, where as some things Pony will just never understand fully.
Cheers.
@fizalpher 100% agreed
@fizalpher I'd say Pony was too heavily fine tuned so loras didn't work so well. Like if you prompted for realism, it was almost deterministic with the style. And creating realistic loras always converged to the weird 3d aesthetic.
With NoobAI, concepts doesn't converge so well to one path. That makes NoobAI feel more random and chaotic but same time it provides more space to express different concepts.
Maybe I'm wrong, so ignore this comment if I am:
Can you please also let the Epsilon-pred 1.0-Version on-site generation enabled for Illustrious loras again?
(Or any other Epsilon version that you judge to be the best)
IIrc it worked great with Illustrious loras.
Since your Epsilon worked great for Illustrious, you could leave it enabled with the Illustrious Base model tag and the V-Pred version stays with the SDXL 1.0 Base model tag, since it seems to not be compatible with Illustrious loras anymore.
Also, thank you for sharing these checkpoints, they're indeed amazing.
我觉得让人比较有挫败感的是，即便是照抄例图的提示以后以后简单改改，有可能图片质量都会下滑，我已经按照用户手册推荐的设置去做，但仍然很难稳定出图质量。
This model made me realize how stubborn Pony was.
The learning ability is amazing.
Like that genius guy at the school preparing to the exam with merely checking the course book.
Up主, can i fine-tune this model to be more nsfw focused, i want a gooner version.
Too lazy to do that though, just asking.
somehow it's getting worse each update, except 0.7
i got skill issue with prompting with this model for some reason
我为 NoobAI-XL v-pre 0.6 版做了一个画师串画风测试excel，下载地址：https://aliang-rec.icu/GOODIES/NoobAI/300%E7%94%BB%E9%A3%8E-NoobAI-V-Pred-0.6%E6%B5%8B%E8%AF%95.xlsx
文件较大（1.38G）无法在线预览，请下载到本地打开
画风串来源为：https://docs.qq.com/sheet/DZWZMemxNZkpVR0VB?tab=BB08J2
I did an artist string style test excel for NoobAI-XL v-pre version 0.6, download address: https://aliang-rec.icu/GOODIES/NoobAI/300%E7%94%BB%E9%A3%8E-NoobAI-V-Pred-0.6%E6%B5%8B%E8%AF%95.xlsx
If the file size (1.38G) is large and cannot be previewed online, please download it to your local computer to open it
The source of the painting style string is: https://docs.qq.com/sheet/DZWZMemxNZkpVR0VB?tab=BB08J2
What causes characters to have blue or white skin all of a sudden? This seems to be a common issue across my generations. (I should say that I really like the model so far, and that this comment isn't meant to be mean! Just trying to figure out what's going on <3)
I've learned a lot of interesting stuff from your documents here. I'll be running some trainings and tests using your settings with flux, and see how they turn out.
"This image generation model [...] leverages full Danbooru and e621 datasets with native tags [...]"
"Negative Prompt:
[...] mammal, anthro, furry, ambiguous form, feral, semi-anthro"
I'm genuinely confused. Isn't e621 mostly furry art? Also, if the boorus use similar tags with different definitions, aren't they going to conflict with each other? Like "floating_hand" on Danbooru and "floating_hands" on E621, they are NOT the same thing. And if the point of the model is to generate anime humans/kemonomimis/monster folk and NOT furries (like the negative prompt base suggests), what was the point of training on e621?
Hello!
What kind of LoRA can i use with IL models? XL or Pony based LoRA will work or not?
Will there be a V-Pred version that works with A1111 without having to use the dev branch?
I think it's a bad call to remove its source model illustrious. Firstly, there's the issue that user-created LoRA category assignments become ambiguous, and secondly, there's a question of reasonableness.
I believe that erasing illustrious as its source stems from an underlying desire to have a new 'noobai' category because the LoRA isn't quite compatible.
While 'illustrious' and 'pony' were given unique categories because they each had their own distinctive features from the start, although this model is excellent, I don't think it has characteristics significant enough to be classified into a separate category within the already saturated categories of Civitai.
Also, It's even more puzzling to use 'sdxl' as the base model instead of 'ilxl', because a LoRA trained on an 'sdxl' base is much less applicable than one trained on an 'ilxl' base. Even if pony-based models have diversified, they still register posts under the same pony category and simply note in the comments that a specific model was used; they don't register posts under the 'sdxl' category just because the pony is based on 'sdxl'.
Of course, noobai incorporates a level of fine-tuning that's on a completely different scale compared to the branches of the pony model, but I still think the argument remains unchanged.
Can this model use with P-Adapter?
why was the v-pred version with its acid lighting left on the site for generation? The epsilon version was much simpler !
My tests show this v-pred v0.6 is more reliable than actual NovelAI at generating baseline txt2img context.
The majority of tags when put into NAI after a certain token count tend to be ignored or do nothing, while this model seems to continue well beyond that point, and does a great job at distributing fractional attention to various elements. Along with that, it handles many high complexity situations that base Illustrious XL turns into garbled noise or chaos with flying colors. The baseline quality may not quite be there YET for v0.6, but damn it's got some MAD context and the quality IS comparable with the correct tokens.
It's not as complex as say a full explanation that Flux can handle, but it only takes a FRACTION of the time of Flux, and with the right tags the quality IS comparable to some of the best models out there, albeit a bit rigid in that state for now.
Img2Img and Inpainting though, NAI has their own inpainting models. They are actually something else.
Food for thought.
Until V0.7 releases I'll probably just be playing with this one over NAI in many instances, then I'll probably just switch to v0.7.
Why do all the actual good generations have their prompts hidden? Does anyone know any good artist tags for anime?
經過三天的使用，這模型實在太強大
PONY XL的幾個微調版本都被我丟進垃圾桶了, 正準備把我所有的角色服裝LORA重新訓練成v-pred的版本
以v-pred LORA訓練下出來, 多概念污染影響幾乎為0, 我昨天就試了一個4個角色和24套服裝(每套服裝又有3-6個分件)訓練的LORA, 以NOOBAI為基底仍然可以還原出接近100分的服裝, 而我以前在PONY XL和其他SDXL上都很容易出現同為dress 類即使有獨立名稱及編號,仍然會混合在一起的情況發生
PONY已經很久沒有更新了,未來也比較黑暗,一直都希望找到更好的模型....noob ai真的值得一試
雖然我是使用Obsession微調版本 , 但測試兩者的LORA使用上差異不大
期待之後的版本更新, 如果能改善背景的細節度就更好了, 感謝大佬們的無私貢獻
至於有人說有穩定性問題,我倒是沒有遇到,可能是因為我有專門訓練LORA去配合,可以很穩定的出圖,手有9成都是好的
with noobai .can i take style from a series?
为啥我只能生成马赛克，我检查了一下还是没找到问题在哪……
when will vpred 1 come out
the example prompts with like 7 artist tags seem very strange, how do you guys come up with these prompts? just try a bunch of random artists and see what comes out?
does SMEA not work with vpred? such a good sampler but it seems to break that version. :(
I was using Forge and moved to reForge just to use this model. I've tried working with the configs and everything, but I have not been able to make this model work. I've been working on this on and off since the day v0.5 came out and have run out of ideas. Can someone help me get this model to work and/or help me troubleshoot?
Epsilon-pred 0.5-version could generate it without any problem, but Epsilon-pred 1-version produced only noise...
Details
Files
noobaiXLNAIXL_vPred06Version.safetensors
Mirrors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
checkpoint-e0_s14000.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobvpred06.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobai-xl-vpred-v0.6.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
noobaiXLNAIXL_vPred06Version.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.