comfyui_meituan_image

<div align="center">

🐱 ComfyUI Meituan Image (LongCat)

ComfyUI integration for Meituan's LongCat-Image model - High-quality bilingual text-to-image generation and image editing

English | 中文

</div>

---

English

📖 Overview

This ComfyUI custom node package provides seamless integration with Meituan's LongCat-Image, a state-of-the-art open-source bilingual (Chinese-English) foundation model for image generation and editing.

✨ Features

| Feature | Description | |---------|-------------| | 🎨 Text-to-Image | Generate high-quality images from text prompts with excellent Chinese text rendering | | ✏️ Image Edit | Edit images using natural language instructions | | ⚡ Performance Options | CPU offload support and optional SageAttention acceleration | | 🔧 Auto Model Detection | Automatically detects LongCat models in your models directory | | 📥 Auto Model Download | Automatically downloads models from HuggingFace if not found locally | | 🔄 Auto Dependency Install | Automatically installs longcat-image package if not present |

📦 Nodes Included

1. LongCat Model Loader - Load LongCat-Image, LongCat-Image-Dev, or LongCat-Image-Edit models
2. LongCat Text to Image - Generate images from text prompts
3. LongCat Image Edit - Single image editing with natural language

🚀 Installation

Prerequisites

• ComfyUI installed and working
• Python 3.10+
• CUDA-capable GPU with at least 17GB VRAM (with CPU offload) or 24GB+ (without)

Step 1: Clone the Repository

bash cd ComfyUI/custom_nodes git clone https://github.com/your-username/comfyui_meituan_image.git

Step 2: Install Dependencies

bash cd comfyui_meituan_image pip install -r requirements.txt

Or install LongCat-Image directly:

bash pip install git+https://github.com/meituan-longcat/LongCat-Image.git@main

Step 3: Download Models (Optional - Auto Download Available)

💡 Note: Starting from v1.1, models are automatically downloaded when you first use a node. You can skip this step if you prefer auto-download.

Model Storage Location:

ComfyUI/
└── models/
    └── diffusion_models/
        ├── LongCat-Image/          # Text-to-Image model
        ├── LongCat-Image-Dev/      # Dev model (faster)
        └── LongCat-Image-Edit/     # Image Editing model

Available Models:

| Model | HuggingFace Repo | Description | |-------|------------------|-------------| | LongCat-Image | meituan/LongCat-Image | Full text-to-image model (50 steps) | | LongCat-Image-Dev | meituan/LongCat-Image-Dev | Faster model (28 steps) | | LongCat-Image-Edit | meituan/LongCat-Image-Edit | Image editing model |

Manual Download (if needed):

pip install "huggingface_hub[cli]"

# Text-to-Image model
huggingface-cli download meituan/LongCat-Image --local-dir ComfyUI/models/diffusion_models/LongCat-Image

# Dev model (faster, 28 steps)
huggingface-cli download meituan/LongCat-Image-Dev --local-dir ComfyUI/models/diffusion_models/LongCat-Image-Dev

# Image Editing model
huggingface-cli download meituan/LongCat-Image-Edit --local-dir ComfyUI/models/diffusion_models/LongCat-Image-Edit

📝 Usage

Text-to-Image Generation

1. Add LongCat Model Loader node and select LongCat-Image or LongCat-Image-Dev
2. Connect to LongCat Text to Image node
3. Enter your prompt (supports both Chinese and English)
4. For text rendering, enclose text in quotes: "你好世界"

Image Editing

1. Add LongCat Model Loader and select LongCat-Image-Edit
2. Connect to LongCat Image Edit node
3. Load your source image and enter editing instructions

⚙️ Node Parameters

LongCat Model Loader

| Parameter | Description | Default | |-----------|-------------|---------| | model_name | Select from detected models | Auto | | custom_model_path | Manual path override | - | | dtype | Precision: bfloat16/float16/float32 | bfloat16 | | enable_cpu_offload | Enable to save VRAM (~17GB needed) | true | | attention_backend | default or sage (requires sageattention) | default |

LongCat Text to Image

| Parameter | Description | Default | |-----------|-------------|---------| | prompt | Image description (Chinese/English) | - | | negative_prompt | What to avoid | - | | width/height | Output dimensions | 1344x768 | | steps | Inference steps | 50 | | guidance_scale | CFG scale | 4.5 | | seed | Random seed | 43 | | batch_size | Number of images | 1 |

LongCat Image Edit

| Parameter | Description | Default | |-----------|-------------|---------| | image | Source image | Required | | prompt | Edit instructions | - | | steps | Inference steps | 50 | | guidance_scale | CFG scale | 4.5 |

📂 Example Workflows

Example workflows are available in the example/ folder:

• example_workflow_t2i.json - Text-to-Image
• example_workflow_edit.json - Image Edit

🔗 Related Links

• LongCat-Image GitHub
• LongCat-Image on Hugging Face
• LongCat Official App
• ComfyUI

---

中文

📖 概述

这是美团 LongCat-Image 长猫图像模型 的 ComfyUI 自定义节点包。LongCat-Image 是一个开源双语（中英文）图像生成与编辑基础模型。

✨ 功能特性

| 功能 | 说明 | |------|------| | 🎨 文生图 | 从文本生成高质量图像，出色的中文渲染能力 | | ✏️ 图像编辑 | 使用自然语言编辑图像 | | ⚡ 性能优化 | 支持 CPU 卸载和 SageAttention 加速 | | 🔧 自动识别 | 自动检测模型目录中的 LongCat 模型 | | 📥 自动下载模型 | 如果本地没有模型，自动从 HuggingFace 下载 | | 🔄 自动安装依赖 | 如果没有安装 longcat-image 包，自动安装 |

📦 包含节点

1. LongCat Model Loader - 加载 LongCat 模型
2. LongCat Text to Image - 文生图节点
3. LongCat Image Edit - 图像编辑节点

🚀 安装方法

环境要求

• 已安装 ComfyUI
• Python 3.10+
• CUDA 显卡，至少 17GB 显存（开启 CPU 卸载）或 24GB+（不卸载）

第一步：克隆仓库

bash cd ComfyUI/custom_nodes git clone https://github.com/your-username/comfyui_meituan_image.git

第二步：安装依赖

bash cd comfyui_meituan_image pip install -r requirements.txt

第三步：下载模型（可选 - 支持自动下载）

💡 提示： 从 v1.1 版本开始，模型会在首次使用时自动下载。如果你希望自动下载，可以跳过这一步。

模型存放位置：

ComfyUI/
└── models/
    └── diffusion_models/
        ├── LongCat-Image/          # 文生图模型
        ├── LongCat-Image-Dev/      # 开发版模型（更快）
        └── LongCat-Image-Edit/     # 图像编辑模型

可用模型：

| 模型 | HuggingFace 仓库 | 说明 | |------|------------------|------| | LongCat-Image | meituan/LongCat-Image | 完整文生图模型（50步） | | LongCat-Image-Dev | meituan/LongCat-Image-Dev | 快速版模型（28步） | | LongCat-Image-Edit | meituan/LongCat-Image-Edit | 图像编辑模型 |

手动下载（可选）：

pip install "huggingface_hub[cli]"

# 文生图模型
huggingface-cli download meituan/LongCat-Image --local-dir ComfyUI/models/diffusion_models/LongCat-Image

# 开发版模型（更快，28步）
huggingface-cli download meituan/LongCat-Image-Dev --local-dir ComfyUI/models/diffusion_models/LongCat-Image-Dev

# 图像编辑模型
huggingface-cli download meituan/LongCat-Image-Edit --local-dir ComfyUI/models/diffusion_models/LongCat-Image-Edit

第四步：重启 ComfyUI

📝 使用方法

文生图

1. 添加 LongCat Model Loader 节点，选择 LongCat-Image
2. 连接到 LongCat Text to Image 节点
3. 输入提示词（支持中英文）
4. 渲染文字时用引号包裹："你好世界"

图像编辑

1. 添加 LongCat Model Loader，选择 LongCat-Image-Edit
2. 连接到 LongCat Image Edit 节点
3. 加载源图像并输入编辑指令

📄 许可证

本项目基于 Apache 2.0 许可证 开源。

🙏 致谢

• 美团长猫团队 - LongCat-Image 模型
• ComfyUI - 强大的图像生成工作流平台