comfyui-api-image-edit 1.0.2

A powerful ComfyUI custom node for API-based image editing with multiple provider support

ComfyUI API Image Edit

中文说明

强大的ComfyUI API图像编辑自定义节点，支持多个API提供商。通过ModelScope、OpenRouter、OpenAI、Google Gemini和PixelWords等API实现文本生图、单图编辑、多图合成和多轮对话编辑功能，具备动态模型加载、智能上下文管理和安全密钥管理能力。

English Description

A powerful ComfyUI custom node for API-based image editing with multiple provider support. Features text-to-image generation, single image editing, multi-image composition, and multi-turn conversation editing through various APIs with dynamic model loading, intelligent context management, and secure key management.

🚀 Features

多API提供商支持

• ModelScope - 千问图像编辑模型
• OpenRouter - 多种视觉模型访问
• OpenAI - GPT-4 Vision 和 DALL-E 模型
• Google Gemini - Gemini 2.0/2.5 Flash 图像生成
• PixelWords - API服务商，支持Gemini 2.5 Flash Image和Gemini 2.5 Flash Image HD

灵活的图像输入

• 文本生图 - 仅通过文字描述创建图像
• 单图编辑 - 用自然语言编辑现有图像
• 多图合成 - 合并最多4张图像进行复杂操作
• 可选图像端口 - 所有图像输入（image1-4）都是可选的

高级功能

• 动态模型加载 - 自动获取每个提供商的可用模型
• API密钥管理 - 安全的本地存储和提供商切换
• 多语言支持 - 支持中文和英文提示词
• 多轮对话模式 - 基于官方Gemini API的连续图像编辑
• 智能上下文 - 自动保持对话历史和图像状态
• 错误处理 - 强大的回退机制和清晰的错误消息

Multi-Provider Support

• ModelScope - Qwen image editing models
• OpenRouter - Access to multiple vision models
• OpenAI - GPT-4 Vision and DALL-E models
• Google Gemini - Gemini 2.0/2.5 Flash image generation
• PixelWords - API service provider supporting Gemini 2.5 Flash Image and Gemini 2.5 Flash Image HD

Flexible Image Input

• Text-to-Image Generation - Create images from text descriptions only
• Single Image Editing - Edit existing images with natural language
• Multi-Image Composition - Combine up to 4 images for complex operations
• Optional Image Ports - All image inputs (image1-4) are optional

Advanced Features

• Dynamic Model Loading - Automatically fetch available models for each provider
• API Key Management - Secure local storage with provider switching
• Multi-Language Support - Supports Chinese and English prompts
• Multi-Turn Conversation - Continuous image editing based on official Gemini API
• Intelligent Context - Automatically maintains conversation history and image state
• Error Handling - Robust fallback mechanisms and clear error messages

📦 安装方法

方式1: ComfyUI Manager (推荐)

1. 通过ComfyUI Manager安装
2. 搜索 "ComfyUI API Image Edit"
3. 点击安装

方式2: 手动安装

cd ComfyUI/custom_nodes
git clone https://github.com/aiaiaikkk/comfyui-api-image-edit.git
cd comfyui-api-image-edit
pip install -r requirements.txt

方式3: PyPI安装

pip install comfyui-api-image-edit

📦 Installation

Option 1: ComfyUI Manager (Recommended)

1. Install through ComfyUI Manager
2. Search for "ComfyUI API Image Edit"
3. Click Install

Option 2: Manual Installation

cd ComfyUI/custom_nodes
git clone https://github.com/aiaiaikkk/comfyui-api-image-edit.git
cd comfyui-api-image-edit
pip install -r requirements.txt

Option 3: PyPI Installation

pip install comfyui-api-image-edit

🔧 Dependencies

• requests>=2.25.1 - HTTP API calls
• Pillow>=8.0.0 - Image processing
• numpy>=1.19.0 - Array operations
• torch>=1.9.0 - ComfyUI tensor operations
• google-genai>=1.32.0 - Google Gemini API support

🎯 使用方法

基本设置

1. 将"API Image Edit"节点添加到ComfyUI工作流
2. 选择你喜欢的API提供商
3. 输入你的API密钥
4. 从下拉菜单中选择模型
5. 编写你的提示词

生成模式

文本生图

• 不要连接任何图像到输入端口
• 在提示词中写下你的描述
• 节点将根据文本生成新图像

单图编辑

• 将图像连接到 image1 端口
• 在提示词中描述你想要的更改
• 示例："将背景更改为日落场景"

多图合成

• 将2-4张图像连接到 image1, image2, image3, image4 端口
• 在提示词中描述合成方式
• 示例："将这些图像合并为拼贴画" 或 "用image2中的脸替换image1中的脸"

多轮对话编辑 🆕

• 连续编辑模式：保持图像编辑的连续性，无需重新上传图像
• 使用方法：
  1. 第一轮：连接图像到 image1，输入提示词如"让这个人笑"
  2. 第二轮：断开 image1 连接，启用 use_last_image，输入"戴上红帽子"
  3. 第三轮：继续输入"改变背景为花园"，系统会基于前面结果继续编辑
• 控制参数：
  • conversation_mode: 启用多轮对话模式（默认开启）
  • use_last_image: 使用对话历史中的最后一张图像
  • reset_conversation: 重置对话开始新的编辑序列

智能工作流程

输入图片 → "笑" → 生成笑脸
    ↓
断开连接 + use_last_image → "戴花" → 在笑脸基础上戴花
    ↓
继续对话 → "海滩背景" → 在戴花笑脸基础上换背景

🎯 Usage

Basic Setup

1. Add the "API Image Edit" node to your ComfyUI workflow
2. Select your preferred API provider
3. Enter your API key
4. Choose a model from the dropdown
5. Write your prompt

Generation Modes

Text-to-Image

• Don't connect any images to the input ports
• Write your description in the prompt
• The node will generate new images based on your text

Single Image Editing

• Connect an image to image1 port
• Describe the changes you want in the prompt
• Example: "Change the background to a sunset scene"

Multi-Image Composition

• Connect 2-4 images to image1, image2, image3, image4 ports
• Describe the composition in the prompt
• Example: "Combine these images into a collage" or "Replace the face in image1 with the face from image2"

Multi-Turn Conversation Editing 🆕

• Continuous Editing Mode: Maintains continuity in image editing without re-uploading images
• How to Use:
  1. First Turn: Connect image to image1, input prompt like "make this person smile"
  2. Second Turn: Disconnect image1, enable use_last_image, input "add a red hat"
  3. Third Turn: Continue with "change background to garden", system will edit based on previous results
• Control Parameters:
  • conversation_mode: Enable multi-turn conversation mode (default: enabled)
  • use_last_image: Use the last image from conversation history
  • reset_conversation: Reset conversation to start a new editing sequence

Smart Workflow

Input Image → "smile" → Generate Smiling Face
    ↓
Disconnect + use_last_image → "add flower" → Add flower to smiling face
    ↓
Continue conversation → "beach background" → Change background while keeping flower and smile

API提供商配置

ModelScope

• 从 ModelScope 获取API密钥
• 支持千问图像编辑模型
• 最适合中文提示词

OpenRouter

• 从 OpenRouter 获取API密钥
• 通过单个API访问多个提供商
• 支持各种视觉模型

OpenAI

• 从 OpenAI 获取API密钥
• 支持GPT-4 Vision和DALL-E模型
• 高质量图像生成

Google Gemini

• 从 Google AI Studio 获取API密钥
• 支持最新的Gemini 2.0 Flash模型
• 高级多模态能力

PixelWords

• API服务商平台
• 支持Gemini 2.5 Flash Image和Gemini 2.5 Flash Image HD
• 通过第三方API访问Google图像生成模型

API Provider Configuration

ModelScope

• Get API key from ModelScope
• Supports Qwen image editing models
• Best for Chinese language prompts

OpenRouter

• Get API key from OpenRouter
• Access to multiple providers through one API
• Supports various vision models

OpenAI

• Get API key from OpenAI
• Supports GPT-4 Vision and DALL-E models
• High quality image generation

Google Gemini

• Get API key from Google AI Studio
• Supports latest Gemini 2.0 Flash models
• Advanced multimodal capabilities

PixelWords

• API service platform
• Supports Gemini 2.5 Flash Image and Gemini 2.5 Flash Image HD
• Third-party API access to Google's image generation models

🛠️ Development

Building from Source

git clone https://github.com/aiaiaikkk/comfyui-api-image-edit.git
cd comfyui-api-image-edit
pip install -e .

Running Tests

pip install -e .[dev]
pytest

📝 Configuration

Node Parameters

必需参数 / Required Parameters

• API Provider - Select from available providers / 选择API提供商
• API Key - Your API key (stored locally and securely) / API密钥（本地安全存储）
• Model - Auto-populated based on provider selection / 基于提供商自动填充的模型选择
• Prompt - Describe your desired image or edits / 描述您想要的图像或编辑

图像输入 / Image Inputs

• Image1-4 - Optional image inputs for editing/composition / 用于编辑/合成的可选图像输入

多轮对话控制 / Multi-Turn Conversation Controls 🆕

• conversation_mode - Enable/disable conversation mode / 启用/禁用对话模式 (默认: True)
• use_last_image - Use last image from history / 使用历史中的最后一张图像 (默认: False)
• reset_conversation - Reset current conversation / 重置当前对话 (默认: False)

高级设置 / Advanced Settings

• Strength - Editing intensity (0.0-2.0) / 编辑强度
• Guidance Scale - Generation guidance (1.0-20.0) / 生成引导系数
• Steps - Sampling steps (1-100) / 采样步数
• Seed - Random seed for reproducibility / 可重现性的随机种子
• Negative Prompt - What to avoid in generation / 生成中要避免的内容
• Watermark - Add watermark to generated images / 为生成图像添加水印
• Refresh Models - Update model list from API / 从API更新模型列表

🔒 Privacy & Security

• API keys are stored locally in browser localStorage
• No data is sent to third parties except chosen API providers
• All processing happens through official API endpoints
• Supports proxy configurations for enhanced privacy

🐛 Troubleshooting

Common Issues

1. "API key is required" - Enter a valid API key for selected provider
2. "Model not available" - Click refresh models or switch providers
3. "Import Error" - Install missing dependencies with pip install -r requirements.txt
4. Network errors - Check internet connection and API key validity

Error Messages

• Clear error messages guide you to solutions
• Automatic fallback to REST APIs when SDK unavailable
• Detailed logging for debugging

多轮对话常见问题 / Multi-Turn Conversation FAQ 🆕

Q: 如何实现连续图像编辑？ / How to achieve continuous image editing? A: 第一轮连接图像，之后断开连接并启用 use_last_image，系统会自动使用上一轮结果。 First turn connect image, then disconnect and enable use_last_image, system will use previous results automatically.

Q: 什么时候需要重置对话？ / When should I reset conversation? A: 开始全新编辑任务时启用 reset_conversation，清除历史记录。 Enable reset_conversation when starting a completely new editing task to clear history.

Q: 支持哪些API提供商的多轮对话？ / Which providers support multi-turn conversation? A: 目前完整支持 Google Gemini，其他提供商的支持正在开发中。 Currently full support for Google Gemini, support for other providers is under development.

Q: 对话历史会占用多少内存？ / How much memory does conversation history use? A: 自动限制最多20条消息和10张图像，避免内存溢出。 Automatically limits to 20 messages and 10 images maximum to prevent memory overflow.

📄 License

MIT License - see LICENSE file for details

🤝 Contributing

Contributions welcome! Please feel free to submit pull requests or open issues.

🔗 Links

• GitHub Repository
• Issue Tracker
• ComfyUI

⭐ Support

If you find this project helpful, please give it a star on GitHub!