<h1 id="model-introduction">Model Introduction</h1>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.<h1 id="important-notice">⚠️ IMPORTANT NOTICE ⚠️</h1><h2 id="this-model-works-different-from-eps-models!">THIS MODEL WORKS DIFFERENT FROM EPS MODELS!</h2><h2 id="please-read-the-guide-carefully!">PLEASE READ THE GUIDE CAREFULLY!</h2><h2 id="model-details">Model Details</h2><ul><li>Developed by: <a target="_blank" rel="ugc" href="https://huggingface.co/Laxhar">Laxhar Lab</a></li><li>Model Type: Diffusion-based text-to-image generative model</li><li>Fine-tuned from: Laxhar/noobai-XL_v1.0</li><li>Sponsored by from:<a target="_blank" rel="ugc" href="https://cloud.lanyun.net/">Lanyun Cloud</a><a target="_blank" rel="ugc" href="https://www.seaart.ai/">Civitai &amp; Seaart</a></li><li>Collaborative testing:<a target="_blank" rel="ugc" href="www.liblib.art">LiblibAI</a><a target="_blank" rel="ugc" href="https://www.nieta.ai/">Nieta</a></li></ul><hr /><h1 id="how-to-use-the-model.">How to Use the Model.</h1><h2 id="guidebook-for-noobai-xl:">Guidebook for NoobAI XL:</h2>ENG:<a target="_blank" rel="ugc" href="https://civarchive.com/articles/8962">https://civarchive.com/articles/8962</a>CHS:<a target="_blank" rel="ugc" href="https://fcnk27d6mpa5.feishu.cn/wiki/S8Z4wy7fSiePNRksiBXcyrUenOh">https://fcnk27d6mpa5.feishu.cn/wiki/S8Z4wy7fSiePNRksiBXcyrUenOh</a><h2 id="recommended-lora-list-for-noobai-xl:">Recommended LoRa List for NoobAI XL:</h2><a target="_blank" rel="ugc" href="https://fcnk27d6mpa5.feishu.cn/wiki/IBVGwvVGViazLYkMgVEcvbklnge">https://fcnk27d6mpa5.feishu.cn/wiki/IBVGwvVGViazLYkMgVEcvbklnge</a><h2 id="method-i:-reforge">Method I: <a target="_blank" rel="ugc" href="https://github.com/Panchovix/stable-diffusion-webui-reForge/">reForge</a></h2><ol><li>(If you haven't installed reForge) Install reForge by following the instructions in the repository;</li><li>Launch WebUI and use the model as usual!</li></ol><h2 id="method-ii:-comfyui">Method II: <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI">ComfyUI</a></h2>SAMLPLE with NODES<a target="_blank" rel="ugc" href="https://huggingface.co/Laxhar/noobai-XL-Vpred-0.5/blob/main/comfy_ui_workflow_sample.png">comfy_ui_workflow_sample</a><h2 id="method-iii:-webui">Method III: <a target="_blank" rel="ugc" href="https://github.com/AUTOMATIC1111/stable-diffusion-webui">WebUI</a></h2>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 simp2.Switch to <code>dev</code> branch:<pre><code>git switch dev
</code></pre>3. Pull latest updates:<pre><code>git pull
</code></pre>4. Launch WebUI and use the model as usual!<h2 id="method-iv:-diffusers">Method IV: <a target="_blank" rel="ugc" href="https://huggingface.co/docs/diffusers/en/index">Diffusers</a></h2><pre><code>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")
</code></pre>Note: Please make sure Git is installed and environment is properly configured on your machine.<hr /><h1 id="recommended-settings">Recommended Settings</h1><h2 id="parameters">Parameters</h2><ul><li>CFG: 4 ~ 5</li><li>Steps: 28 ~ 35</li><li>Sampling Method: Euler (⚠️ Other samplers will not work properly)</li><li>Resolution: Total area around 1024x1024. Best to choose from: 768x1344, 832x1216, 896x1152, 1024x1024, 1152x896, 1216x832, 1344x768</li></ul><h2 id="prompts">Prompts</h2><ul><li>Prompt Prefix:</li></ul><pre><code>masterpiece, best quality, newest, absurdres, highres, safe,
</code></pre><ul><li>Negative Prompt:</li></ul><pre><code>nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro
</code></pre><h1 id="usage-guidelines">Usage Guidelines</h1><h2 id="caption">Caption</h2><pre><code>&lt;1girl/1boy/1other/...&gt;, &lt;character&gt;, &lt;series&gt;, &lt;artists&gt;, &lt;special tags&gt;, &lt;general tags&gt;, &lt;other tags&gt;
</code></pre><h2 id="quality-tags">Quality Tags</h2>For quality tags, we evaluated image popularity through the following process:<ul><li>Data normalization based on various sources and ratings.</li><li>Application of time-based decay coefficients according to date recency.</li><li>Ranking of images within the entire dataset based on this processing.</li></ul>Our ultimate goal is to ensure that quality tags effectively track user preferences in recent years.Percentile RangeQuality Tags&gt; 95thmasterpiece&gt; 85th, &lt;= 95thbest quality&gt; 60th, &lt;= 85thgood quality&gt; 30th, &lt;= 60thnormal quality&lt;= 30thworst quality<h2 id="aesthetic-tags">Aesthetic Tags</h2>TagDescriptionvery awaTop 5% of images in terms of aesthetic score by <a target="_blank" rel="ugc" href="https://huggingface.co/Eugeoter/waifu-scorer-v4-beta">waifu-scorer</a>worst aestheticAll the bottom 5% of images in terms of aesthetic score by <a target="_blank" rel="ugc" href="https://huggingface.co/Eugeoter/waifu-scorer-v4-beta">waifu-scorer</a> and <a target="_blank" rel="ugc" href="https://huggingface.co/shadowlilac/aesthetic-shadow-v2">aesthetic-shadow-v2</a>......<h2 id="date-tags">Date Tags</h2>There are two types of date tags: year tags and period tags. For year tags, use <code>year xxxx</code> format, i.e., <code>year 2021</code>. For period tags, please refer to the following table:Year RangePeriod tag2005-2010old2011-2014early2014-2017mid2018-2020recent2021-2024newest<h2 id="dataset">Dataset</h2><ul><li>The latest Danbooru images up to the training date (approximately before 2024-10-23)</li><li>E621 images <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/NebulaeWis/e621-2024-webp-4Mpixel">e621-2024-webp-4Mpixel</a> dataset on Hugging Face</li></ul>Communication<ul><li>QQ Groups:<ul><li>427280545</li><li>677964513</li><li>852429527</li><li>914818692</li><li>635772191</li><li>870086562</li></ul></li><li>Discord: <a target="_blank" rel="ugc" href="https://discord.com/invite/DKnFjKEEvH">Laxhar Dream Lab SDXL NOOB</a></li></ul>How to train a LoRA on v-pred SDXL modelA tutorial is intended for LoRA trainers based on sd-scripts.article link: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/8723">https://civarchive.com/articles/8723</a>Utility ToolLaxhar 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: <a target="_blank" rel="ugc" href="https://civarchive.com/models/929685">https://civarchive.com/models/929685</a><h1 id="model-license">Model License</h1>This model's license inherits from <a target="_blank" rel="ugc" href="https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0">https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0</a> 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.<h2 id="i.-usage-restrictions">I. Usage Restrictions</h2><ul><li>Prohibited use for harmful, malicious, or illegal activities, including but not limited to harassment, threats, and spreading misinformation.</li><li>Prohibited generation of unethical or offensive content.</li><li>Prohibited violation of laws and regulations in the user's jurisdiction.</li></ul><h2 id="ii.-commercial-prohibition">II. Commercial Prohibition</h2>We prohibit any form of commercialization, including but not limited to monetization or commercial use of the model, derivative models, or model-generated products.<h2 id="iii.-open-source-community">III. Open Source Community</h2>To foster a thriving open-source community,users MUST comply with the following requirements:<ul><li>Open source derivative models, merged models, LoRAs, and products based on the above models.</li><li>Share work details such as synthesis formulas, prompts, and workflows.</li><li>Follow the fair-ai-public-license to ensure derivative works remain open source.</li></ul><h2 id="iv.-disclaimer">IV. Disclaimer</h2>Generated models may produce unexpected or harmful outputs. Users must assume all risks and potential consequences of usage.<h1 id="participants-and-contributors">Participants and Contributors</h1><h2 id="participants">Participants</h2><ul><li>L_A_X: <a target="_blank" rel="ugc" href="https://civarchive.com/user/L_A_X">Civitai</a> | <a target="_blank" rel="ugc" href="http://Liblib.art">Liblib.art</a> | <a target="_blank" rel="ugc" href="https://huggingface.co/LAXMAYDAY">Huggingface</a></li><li>li_li: <a target="_blank" rel="ugc" href="https://civarchive.com/user/li_li">Civitai</a> | <a target="_blank" rel="ugc" href="https://huggingface.co/heziiiii">Huggingface</a></li><li>nebulae: <a target="_blank" rel="ugc" href="https://civarchive.com/user/kitarz">Civitai</a> | <a target="_blank" rel="ugc" href="https://huggingface.co/NebulaeWis">Huggingface</a></li><li>Chenkin: <a target="_blank" rel="ugc" href="https://civarchive.com/user/Chenkin">Civitai</a> | <a target="_blank" rel="ugc" href="https://huggingface.co/windsingai">Huggingface</a></li><li>Euge: <a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">Civitai</a> | <a target="_blank" rel="ugc" href="https://huggingface.co/Eugeoter">Huggingface</a> | <a target="_blank" rel="ugc" href="https://github.com/Eugeoter">Github</a></li></ul><h2 id="contributors">Contributors</h2><ul><li>Narugo1992: Thanks to <a target="_blank" rel="ugc" href="https://github.com/narugo1992">narugo1992</a> and the <a target="_blank" rel="ugc" href="https://huggingface.co/deepghs">deepghs</a> team for open-sourcing various training sets, image processing tools, and models.</li><li>Mikubill: Thanks to <a target="_blank" rel="ugc" href="https://github.com/Mikubill">Mikubill</a> for the <a target="_blank" rel="ugc" href="https://github.com/Mikubill/naifu">Naifu</a> trainer.</li><li>Onommai: Thanks to <a target="_blank" rel="ugc" href="https://onomaai.com/">OnommAI</a> for open-sourcing a powerful base model.</li><li>V-Prediction: Thanks to the following individuals for their detailed instructions and experiments.<ul><li>adsfssdf</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/user/bluvoll">bluvoll</a></li><li><a target="_blank" rel="ugc" href="https://github.com/bvhari">bvhari</a></li><li><a target="_blank" rel="ugc" href="https://github.com/catboxanon">catboxanon</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/parsee-mizuhashi">parsee-mizuhashi</a></li><li><a target="_blank" rel="ugc" href="https://github.com/very-aesthetic">very-aesthetic</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/user/momoura">momoura</a></li><li>madmanfourohfour</li></ul></li><li>Community: <a target="_blank" rel="ugc" href="https://civarchive.com/user/aria1th261">aria1th261</a>, <a target="_blank" rel="ugc" href="https://github.com/neggles/neurosis">neggles</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/sdtana">sdtana</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/chewing">chewing</a>, <a target="_blank" rel="ugc" href="https://github.com/irldoggo">irldoggo</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/reoe">reoe</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/kblueleaf">kblueleaf</a>, <a target="_blank" rel="ugc" href="https://github.com/Yidhar">Yidhar</a>, ageless, 白玲可, Creeper, KaerMorh, 吟游诗人, SeASnAkE, <a target="_blank" rel="ugc" href="https://civarchive.com/user/zwh20081">zwh20081</a>, Wenaka⁧~喵, 稀里哗啦, 幸运二副, 昨日の約, 445, <a target="_blank" rel="ugc" href="https://civarchive.com/user/EBIX">EBIX</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/goyishsoyish">Sopp</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/Y_X">Y_X</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/Minthybasis">Minthybasis</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/Rakosz">Rakosz</a>, 孤辰NULL, 汤人烂, 沅月弯刀,David, 年糕特工队，</li></ul>

NoobAI-XL (NAI-XL)

Merry Christmas! NOOBAI XL-VPred 1.0 has been released! The V prediction series has come to a successful end, and what an interesting journey it has been. Who knows, we might have the chance to do this again in the future. With this, our Laxhar Lab's weekly update plan comes to a grand finale!By the way, here are the advantages of this version:1. Fine-tuned with high-quality datasets: We have optimized the model for anatomical accuracy and compositional rationality through meticulous adjustments with high-quality datasets.2. Flexible style combination weights: The model now offers more flexibility in combining different painting styles, with improved robustness when overlaying multiple styles.3. Enhanced utility of quality words: The effectiveness of quality words has become more pronounced in this version.4. A blend of features from the standard and S versions: The color style is vibrant yet less prone to overexposure, combining the best of both worlds.Usage recommendations and future work:1. Use with dynamic CFG plugin: We recommend using the dynamic CFG &amp; CFG Rescale plugin when employing the V prediction model to prevent oversaturation or overly gray images. You can refer to the configuration of 0.2 for the best results.2. Choice of sampling methods: Although NOOBAI XL-VPred 1.0 supports most sampling methods, V prediction does not support the Karras series of sampling. Therefore, we suggest using Euler and DDIM sampling methods for more stable outcomes.3. Ongoing updates and support: We will continue to update VPred 1.0, including the ControlNet model and other plugins. The main model will also be updated irregularly when there are significant improvements (let's see how the DIT effect of NAI4 is and learn from it), so stay tuned!Lastly, I must share a personal recommendation: the recent game "MiSide" is truly a great game. It has been a long time since a solo game has moved me so much. I strongly recommend it. Wishing you all a Merry Christmas. After a year of hard work, it's time to rest. Until we meet again in this world of possibilitiesミ(・・)ミ

Files with SHA256: ea349eeae87ca8d25ba902c93810f7ca83e5c82f920edf12f273af004ae02819