<h3 id="follow-for-more-updates-at-http:discord.cominvitetttgccjbea"><a target="_blank" rel="ugc" href="http://discord.com/invite/TTTGccjbEa">- Follow for more updates at http://discord.com/invite/TTTGccjbEa</a></h3><h3 id="try-model:-huggingface-playground"><a target="_blank" rel="ugc" href="https://huggingface.co/spaces/neta-art/NetaLumina_T2I_Playground">- Try Model: Huggingface Playground</a></h3><h3 id="access-to-more-training-versions"><a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/lu2">- Access to more training versions</a></h3><h2><a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/Neta-Lumina/blob/main/README-ZH.md">- 中文模型说明</a></h2><h3 id="qq-group:-1039442542"><a target="_blank" rel="ugc" href="https://qm.qq.com/q/vGdplsMR6o">- QQ group: 1039442542</a></h3><hr /><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/SLriOamDn8M" start="0"></iframe></div><h2 id="introduction">Introduction</h2>Neta Lumina is a high‑quality anime‑style image‑generation model developed by <a target="_blank" rel="ugc" href="http://Neta.art">Neta.art</a> Lab.Building on the open‑source Lumina‑Image‑2.0 released by the Alpha‑VLLM team at Shanghai AI Laboratory, we fine‑tuned the model with a vast corpus of high‑quality anime images and multilingual tag data. The preliminary result is a compelling model with powerful comprehension and interpretation abilities (thanks to Gemma text encoder), ideal for illustration, posters, storyboards, character design, and more.<h2 id="key-features">Key Features</h2><ul><li>Optimized for diverse creative scenarios such as Furry, Guofeng (traditional‑Chinese aesthetics), pets, etc.</li><li>Wide coverage of characters and styles, from popular to niche concepts. (Still support danbooru tags!)</li><li>Accurate natural‑language understanding with excellent adherence to complex prompts.</li><li>Native multilingual support, with Chinese, English, and Japanese recommended first.</li></ul><h2 id="model-versions">Model Versions</h2>For models in alpha tests, requst access at <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/NetaLumina_Alpha">https://huggingface.co/neta-art/NetaLumina_Alpha</a> if you are interested.Neta-lumina-v1.0Request access at <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/Neta-Lumina">https://huggingface.co/neta-art/Neta-Lumina</a> if you are interested.<ul><li>Official Release: overall best performance</li></ul>Neta-lumina-beta-0624<ul><li>Primary Goal: General knowledge and anime‑style optimization</li><li>Data Set: &gt;13 million anime‑style images</li><li>&gt;46,000 A100 Hours</li></ul><h2 id="how-to-use">How  to  Use                                                                     </h2>Neta Lumina is built on the Lumina2 Diffusion Transformer (DiT) framework, please follow these steps precisely.<h3 id="comfyui">ComfyUI</h3><h3 id="environment-requirements">Environment Requirements</h3>Currently Neta Lumina runs only on ComfyUI:<ul><li>Latest ComfyUI installation</li><li>≥ 8 GB VRAM</li></ul><h3 id="downloads-and-installation">Downloads &amp; Installation</h3>The model provided by Civitai is a three-in-one (te, dit, vae) packaged version, which can be run using the comfyui basic workflow without the need to download Text Encoder and VAE separately.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fb24daca-3cbd-479f-8326-ab1e96f2edfe/width=525/fb24daca-3cbd-479f-8326-ab1e96f2edfe.jpeg" />Original (component) release<ol><li>Neta Lumina-V1.0<ol><li>Hugging Face: <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/Neta-Lumina/blob/main/Unet/neta-lumina-v1.0.safetensors">https://huggingface.co/neta-art/Neta-Lumina/blob/main/Unet/neta-lumina-v1.0.safetensors</a></li><li>Save path: <code>ComfyUI/models/unet/</code></li></ol></li><li>Text Encoder (Gemma-2B)<ol><li>Download link: <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/Neta-Lumina/blob/main/Text%20Encoder/gemma_2_2b_fp16.safetensors">https://huggingface.co/neta-art/Neta-Lumina/blob/main/Text%20Encoder/gemma_2_2b_fp16.safetensors</a></li><li>Save path: <code>ComfyUI/models/text_encoders/</code></li></ol></li><li>VAE Model (16-Channel FLUX VAE)<ol><li>Download link: <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/Neta-Lumina/blob/main/VAE/ae.safetensors">https://huggingface.co/neta-art/Neta-Lumina/blob/main/VAE/ae.safetensors</a></li><li>Save path: <code>ComfyUI/models/vae/</code></li></ol></li></ol><img src="https://cdn-uploads.huggingface.co/production/uploads/655319e00166ff6bd2351948/B69od-a8Rs4gf1LabuKzX.png" alt="image/png" />Workflow: load <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/Neta-Lumina/blob/main/lumina_workflow.json">lumina_workflow.json</a> in ComfyUI.<ul><li>UNETLoader – loads the .pth</li><li>VAELoader – loads ae.safetensors</li><li>CLIPLoader – loads gemma_2_2b_fp16.safetensors</li><li>Text Encoder – connects positive /negative prompts to the sampler</li></ul>Simple merged releaseDownload neta-lumina-v1.0-all-in-one.safetensors,md5sum = dca54fef3c64e942c1a62a741c4f9d8a,you may use ComfyUI’s simple checkpoint loader workflow.<h3 id="recommended-settings">Recommended Settings</h3><ul><li>Sampler: res_multistep</li><li>Scheduler: linear_quadratic</li><li>Steps: 30</li><li>CFG (guidance): 4 – 5.5</li><li>EmptySD3LatentImage resolution: 1024 × 1024, 768 × 1532, 968 × 1322, or &gt;= 1024 </li></ul><h1 id="prompt-book">Prompt Book </h1><h3 id="detailed-prompt-guidelines:-https:civitai.comarticles16274neta-lumina-drawing-model-prompt-guide">Detailed prompt guidelines: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/16274/neta-lumina-drawing-model-prompt-guide">https://civarchive.com/articles/16274/neta-lumina-drawing-model-prompt-guide</a></h3><h1 id="community">Community                                   </h1><h3 id="discord:-https:discord.cominvitetttgccjbea">Discord: <a target="_blank" rel="ugc" href="https://discord.com/invite/TTTGccjbEa">https://discord.com/invite/TTTGccjbEa</a></h3><h3 id="qq-group:-1039442542">QQ group: 1039442542</h3><h1 id="roadmap">Roadmap </h1><h3 id="model">Model</h3><ul><li>Continous base‑model training to raise reasoning capability.</li><li>Aesthetic‑dataset iteration to improve anatomy, background richness, and overall appealness.</li><li>Smarter, more versatile tagging tools to lower the creative barrier.</li></ul><h3 id="ecosystem">Ecosystem</h3><h3 id="lora-training-tutorials-and-components"><a target="_blank" rel="ugc" href="https://docs.google.com/document/d/1fSp8wgraQeeaM58DVbnFoRULyyhPvt3MgY0Qja0Otc0/edit?tab=t.0">LoRA training tutorials and components</a></h3><ul><li>Development of advanced control / style‑consistency features (e.g., <a target="_blank" rel="ugc" href="https://arxiv.org/pdf/2411.15098">Omini Control</a>). <a target="_blank" rel="ugc" href="https://discord.com/invite/TTTGccjbEa">Call for Collaboration</a>!</li></ul><h1 id="license-and-disclaimer">License &amp; Disclaimer</h1><ul><li>Neta Lumina is released under <a target="_blank" rel="ugc" href="https://www.apache.org/licenses/LICENSE-2.0">Apache License 2.0</a></li></ul><h1 id="participants-and-contributors">Participants &amp; Contributors</h1><ul><li>Special thanks to the <a target="_blank" rel="ugc" href="https://github.com/Alpha-VLLM">Alpha‑VLLM</a> team for open‑sourcing Lumina‑Image‑2.0</li><li>Model development: <a target="_blank" rel="ugc" href="http://Neta.art">Neta.art</a> Lab (<a target="_blank" rel="ugc" href="https://civarchive.com/user/neta_art">Civitai</a>), especially li_li</li><li>Partners<ul><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">Hugging Face</a></li></ul></li><li><a target="_blank" rel="ugc" href="https://github.com/narugo1992">narugo1992</a> &amp; <a target="_blank" rel="ugc" href="https://huggingface.co/deepghs">deepghs</a>: open datasets, processing tools, and models</li><li><a target="_blank" rel="ugc" href="https://github.com/Mikubill/naifu">Naifu</a> trainer at <a target="_blank" rel="ugc" href="https://github.com/Mikubill">Mikubill</a></li></ul><h2 id="community-contributors">Community Contributors</h2>Evaluators &amp; developers: <a target="_blank" rel="ugc" href="https://huggingface.co/Second222">二小姐</a>, <a target="_blank" rel="ugc" href="https://github.com/spawner1145">spawner</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/Rnglg2">Rnglg2</a>Other contributors: <a target="_blank" rel="ugc" href="https://www.pixiv.net/users/22433944">沉迷摸鱼</a>, <a target="_blank" rel="ugc" href="https://x.com/poi______1">poi</a>, AshenWitch, <a target="_blank" rel="ugc" href="https://www.pixiv.net/users/15750592">十分无奈</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/ghostlxh">GHOSTLX</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/Wenaka_">wenaka</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/Blueberries_i">iiiiii</a>, <a target="_blank" rel="ugc" href="https://x.com/gaonian2331">年糕特工队</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/NPCde">恩匹希</a>, 奶冻, <a target="_blank" rel="ugc" href="https://civarchive.com/user/mumu520">mumu</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/user/yizyin">yizyin</a>, smile, Yang, 古神, 灵之药, <a target="_blank" rel="ugc" href="https://civarchive.com/user/LyloGummy">LyloGummy</a>, 雪时<h1 id="appendix-and-resources">Appendix &amp; Resources</h1><ul><li>TeaCache: <a target="_blank" rel="ugc" href="https://github.com/spawner1145/CUI-Lumina2-TeaCache">https://github.com/spawner1145/CUI-Lumina2-TeaCache</a></li><li>Advanced samplers &amp; TeaCache guide (by spawner): <a target="_blank" rel="ugc" href="https://docs.qq.com/doc/DZEFKb1ZrZVZiUmxw?nlc=1">https://docs.qq.com/doc/DZEFKb1ZrZVZiUmxw?nlc=1</a></li><li>Neta Lumina ComfyUI Manual (in Chinese): <a target="_blank" rel="ugc" href="https://docs.qq.com/doc/DZEVQZFdtaERPdXVh">https://docs.qq.com/doc/DZEVQZFdtaERPdXVh</a></li></ul><hr />license: otherlicense_name: fair-ai-public-license-1.0-sdlicense_link: <a target="_blank" rel="ugc" href="https://freedevproject.org/faipl-1.0-sd/">https://freedevproject.org/faipl-1.0-sd/</a><hr />

Neta Lumina

<a target="_blank" rel="ugc" href="https://discord.com/invite/TTTGccjbEa">###Call for collaboration</a>Latest versionNeta-lumina-beta-0624<ul><li>First aesthetic experimental public version</li><li>Primary Goal: Enhanced aesthetics, pose accuracy, and scene detail</li><li>Data Set: Hundreds of thousands of handpicked high‑quality anime images (fine‑tuned on the Base)</li></ul>Base ModelRequest access at <a target="_blank" rel="ugc" href="https://huggingface.co/neta-art/NetaLumina_Alpha">https://huggingface.co/neta-art/NetaLumina_Alpha</a> if you are inteseted.<ul><li>Primary Goal: General knowledge and anime‑style optimization</li><li>Data Set: &gt;13 million anime‑style images</li></ul>

Files with SHA256: 2807541b5677bee767ef073c57f3dc95e75a8fc12dba0e7498188123a0fde0c9