Image with ComfyUI

Call a local ComfyUI server to generate or edit images and videos. Four modes:

• T2I (Text → Image) → Z-Image or SD3.5 Medium model
• I2I (Image → Image / Edit) → Qwen Image Edit model
• I2V (Image → Video) → Wan2.2 model

When to Use

• User asks to generate images from text (Chinese: 绘图/生图/画图/生成图片)
• User asks to edit an image (Chinese: 修图/改图/编辑图片/换装/换背景)
• User asks to generate a video from an image + text (Chinese: 图生视频/动画化/生成视频)
• User provides a description and wants visual output

Image-First Conversational Pattern (Image-First Mode)

Detection rules:

1. User sends only an image (no text, no other message in the same turn)