<h2 id="ming-image-0.1-design-text-encoder">Ming-Image-0.1-Design Text Encoder</h2><p>The original text encoder of inclusionAI's Ming-Image-0.1 (Ling-mini-2.0)</p><h3 id="what-it-is">What it is</h3><ul><li><p>The model that reads your prompt for Ming-Image-0.1-Design: Ling-mini-2.0, a mixture-of-experts language model with 16.3B parameters, 1.4B of them active per token</p></li><li><p>w4a8, 12.8 GB: the version that runs beside the int8 Ming model on an 8 GB card with 32 GB of RAM</p></li></ul><h3 id="bigger-versions">Bigger versions</h3><p>Comfy-Org also publishes this encoder in int8 convrot (19.5 GB) and bf16 (36.7 GB): <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Ming-Image">https://huggingface.co/Comfy-Org/Ming-Image</a></p><h3 id="no-abliteration-needed">No abliteration needed</h3><p>Ming-Image-0.1 renders nudity with this original encoder: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2961185.">https://civarchive.com/models/2961185.</a> Explicit sex acts are not in the base model.</p><h3 id="works-with">Works with</h3><ul><li><p>Ming-Image-0.1-Design, whose files are on Civitai here: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2961185">https://civarchive.com/models/2961185</a></p></li><li><p>The Ming-Image-0.1 VAE: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2961357">https://civarchive.com/models/2961357</a></p></li><li><p>Not Ming-Image-0.1-Design-Layer, which has its own encoder files</p></li></ul><h3 id="install">Install</h3><ul><li><p>Put it in ComfyUI/models/text_encoders/ and load it with the Load CLIP node: ComfyUI recognises it from its weights, whatever the type says</p></li><li><p>Needs ComfyUI with native Ming Image support: the main branch from 24 September 2026, or the next release</p></li></ul><h3 id="about-the-images">About the images</h3><p>Rendered with Ming-Image-0.1-Design (int8 convrot, 12 steps, cfg 1, 832x1248) and this encoder. Each prompt tests something the encoder has to get right: text in English and Chinese, counts, placement, several people with their own clothes, posters and app screens.</p><p>中文：Ming-Image-0.1 原版文本编码器（Ling-mini-2.0，MoE 结构，总参数 163 亿，每个 token 激活 14 亿），未经修改，w4a8 版本（12.8 GB），与 Comfy-Org 发布的文件逐字节一致，8G 显存加 32G 内存即可运行。放入 ComfyUI/models/text_encoders/，用 Load CLIP 节点加载。需要原生支持 Ming 的 ComfyUI。MIT 许可。</p><h3 id="license">License</h3><p>MIT, Copyright (c) 2026 inclusionAI.</p>

Ming-Image-0.1-Design Text Encoder (Ling-mini-2.0)

<p>The original Ling-mini-2.0 text encoder of Ming-Image-0.1</p>

mingImage01DesignText_v10.safetensors

v1.0

<p>The Ling-mini-2.0 text encoders that the Ming-Image-0.1-Design models run on, hosted separately so any Ming Image checkpoint - ours or a community finetune - can point at them instead of shipping its own copy.</p>

<p><strong>Originally released by <a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI">inclusionAI</a> on Hugging Face under the MIT license</strong>, and repackaged as single files for ComfyUI by <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Ming-Image">Comfy-Org</a>. All credit for the weights goes to the inclusionAI team.</p>

<h3><strong>Two versions, and they are not interchangeable</strong></h3>
<p>The base checkpoint and the layer-decomposition checkpoint use different encoder weights. Pick the version that matches the model you are running:</p>
<ul>
<li><strong>Design</strong> - for <a target="_blank" rel="ugc" href="https://civarchive.com/models/2961930">Ming Image Design</a> and anything finetuned from it. Available in BF16, int8 and w4a8.</li>
<li><strong>Layer</strong> - for <a target="_blank" rel="ugc" href="https://civarchive.com/models/2962010">Ming Image Design Layer</a>. Available in BF16 and int8.</li>
</ul>
<p>Loading the wrong one does not error cleanly; it produces noise. If you are unsure which you have, match the filename - the Layer weights carry <code>layer</code> in the name.</p>

<h3><strong>Pick one file, not all of them</strong></h3>
<p>Each file is the whole encoder at a different precision, so you need exactly one. BF16 is the reference weights at full size. int8 is the practical default for most setups. w4a8 (Design only) is the smallest and the one to reach for on limited VRAM. Every file is marked optional for that reason - nothing here requires anything else here.</p>

<h3><strong>Where these go</strong></h3>
<p>In ComfyUI, <code>models/text_encoders/</code>. This is a component rather than a runnable model, so there is no Generate button and no on-site generation - if you want to generate with Ming, that is the <a target="_blank" rel="ugc" href="https://civarchive.com/models/2961930">Ming Image Design</a> page, where the encoder is already handled for you.</p>

<h3><strong>Links</strong></h3>
<ul>
<li><strong>Source:</strong> <a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI/Ming-Image-0.1-Design">Ming-Image-0.1-Design</a> and <a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI/Ming-Image-0.1-Design-Layer">Ming-Image-0.1-Design-Layer</a></li>
<li><strong>Repackaged files:</strong> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Ming-Image">Comfy-Org/Ming-Image</a></li>
<li><strong>Code:</strong> <a target="_blank" rel="ugc" href="https://github.com/inclusionAI/Ming-Image">github.com/inclusionAI/Ming-Image</a></li>
</ul>

Ming Image Text Encoders

ming_image_0.1_ling_mini_2.0_bf16.safetensors

ming_image_0.1_ling_mini_2.0_w4a8.safetensors

ming_image_0.1_ling_mini_2.0_int8_convrot.safetensors

Design

Layer

<p>Ming-Image-0.1-Design is a 6B text-to-image model built for visual design work: UI mockups, infographics, posters, slides and anything else where the text has to come out right. It renders legible type as part of the composition instead of the usual generative-model scribble, and it can output RGBA with a transparent background.</p>

<p><strong>Originally released by <a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI/Ming-Image-0.1-Design">inclusionAI</a> on Hugging Face under the MIT license.</strong> All credit for the model goes to the inclusionAI team. Civitai is hosting a mirror so creators can run it on-site without a local install - please head to the original repository for weights, updates and to follow the project directly.</p>

<h3><strong>Built by</strong></h3>
<ul>
<li><a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI">inclusionAI</a> - the Ming-Image series and the Ling-mini-2.0 text encoder it runs on</li>
<li><a target="_blank" rel="ugc" href="https://github.com/inclusionAI/Ming-Image">The Ming-Image repository</a> - reference inference code, prompt rewriting and the layer-decomposition demo</li>
<li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Ming-Image">Comfy-Org</a> - the repackaged single-file weights this mirror is built from</li>
</ul>

<h3><strong>Text-rich design, not just pictures</strong></h3>
<p>The model was trained for complete visual compositions rather than single subjects. Prompts that describe a layout - headings, body copy, callouts, spacing - are what it is built to answer, and it places readable text inside that layout. It sits near the top of the Artificial Analysis UI/UX design leaderboard.</p>

<h3><strong>Transparent backgrounds</strong></h3>
<p>Ming-Image-0.1-Design generates RGBA output with genuine alpha, so a logo, sticker or UI element comes out ready to composite. Upstream recommends prepending one of the RGBA trigger phrases documented in the <a target="_blank" rel="ugc" href="https://github.com/inclusionAI/Ming-Image#transparent-background-generation-tip">transparent-background tip</a> to switch it on.</p>

<h3><strong>Recommended settings</strong></h3>
<ul>
<li><strong>Resolution:</strong> 2048 x 2048 for best quality, 1024 x 1024 for speed. Output is square at the selected bucket.</li>
<li><strong>Steps:</strong> 12.</li>
<li><strong>CFG scale:</strong> 1.0. Negative prompts do nothing at CFG 1.</li>
<li><strong>Precision:</strong> BF16.</li>
</ul>
<p>Upstream also suggests running prompts through a rewriter first - <code>Ling-3.0-flash-VL</code> or a comparable VLM - since the model responds well to long, structured design briefs.</p>

<h3><strong>The Layer model</strong></h3>
<p>The series has a second 6B model, <a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI/Ming-Image-0.1-Design-Layer">Ming-Image-0.1-Design-Layer</a>, which takes a flattened design and splits it back into separate transparent layers you can edit. It is a different checkpoint with its own task and is not part of this model page.</p>

<h3><strong>Links</strong></h3>
<ul>
<li><strong>Source:</strong> <a target="_blank" rel="ugc" href="https://huggingface.co/inclusionAI/Ming-Image-0.1-Design">inclusionAI/Ming-Image-0.1-Design</a> on Hugging Face</li>
<li><strong>Code:</strong> <a target="_blank" rel="ugc" href="https://github.com/inclusionAI/Ming-Image">github.com/inclusionAI/Ming-Image</a></li>
<li><strong>ModelScope:</strong> <a target="_blank" rel="ugc" href="https://www.modelscope.cn/models/inclusionAI/Ming-Image-0.1-Design">inclusionAI/Ming-Image-0.1-Design</a></li>
<li><strong>Demo:</strong> <a target="_blank" rel="ugc" href="https://huggingface.co/spaces/hugging-apps/ming-image-0-1-design-demo">Hugging Face Space</a></li>
<li><strong>Agent skills:</strong> <a target="_blank" rel="ugc" href="https://github.com/inclusionAI/ling-cookbook/tree/main/resources/recommended-skills/ling-ui-design">Ling UI Design</a> and <a target="_blank" rel="ugc" href="https://github.com/inclusionAI/ling-cookbook/tree/main/resources/recommended-skills/image-to-editable-ppt">Image to Editable PPT</a></li>
</ul>

Files with SHA256: 91de4cd0718bec1452b74ff3b0df0dfde280493dab384f0eb4007c18bf932ccd