Ming-Image-0.1-Design Text-to-Image Workflow (ComfyUI)

Ming-Image-0.1-Design Text-to-Image Workflow

A text-to-image workflow for inclusionAI's Ming-Image-0.1-Design in ComfyUI: core nodes only, runs on an 8 GB card.

What you get

• One clean graph in four groups: models, prompt, size, sample and save
• ComfyUI offers to download the three model files if they are missing (Comfy-Org, Hugging Face)
• The reference settings: 12 steps, cfg 1, euler, simple
• A note in the graph with sizes, timings and prompting tips

Files it loads

• ming_image_0.1_design_int8_convrot (6.2 GB) in models/diffusion_models, or the bf16 file (12.3 GB)
• ming_image_0.1_ling_mini_2.0_w4a8 (12.8 GB) in models/text_encoders
• ming_image_vae_bf16 (0.25 GB) in models/vae
• The model files are also on Civitai: https://civitai.com/models/2961185

Size and speed

• 1024 (about 1 megapixel: 1024x1024, 832x1248 or 1248x832): about 22 s per image on an RTX 4070 Laptop 8 GB
• 2048 (2048x2048, the size inclusionAI recommends): about 85 s on the same card

Prompting

Write a plain caption: who or what, what happens, where, the light. Put any text to render in double quotes; English and Chinese both work. At cfg 1 the negative prompt does nothing.

Needs

ComfyUI with native Ming Image support: the main branch from 24 September 2026, or the next release. No custom nodes.

About the images

Made with this workflow, unchanged, at 832x1248. Drag any of them into ComfyUI to load it.

中文：Ming-Image-0.1-Design 的 ComfyUI 文生图工作流，只用核心节点，8G 显存可运行。缺少的模型文件 ComfyUI 会提示下载。12 步，cfg 1，euler，simple；1024 尺寸约 22 秒一张，2048 尺寸约 85 秒。需要原生支持 Ming 的 ComfyUI。

License

Workflow: MIT, noctaluna. Ming-Image-0.1: MIT, Copyright (c) 2026 inclusionAI.