comfyui-workflow-builder

ComfyUI Workflow 构建指南

⚠️ 操作前必须先阅读 comfyui-operation-rules

---

Workflow JSON 结构

workflow = {
    "node_id": {                          # 字符串 ID（"1", "2", ...）
        "class_type": "NodeClassName",     # 节点类型名
        "inputs": {
            "param": value,                # 直接值
            "input": ["other_id", idx]     # 连接: [源节点ID, 输出索引(0-based)]
        }
    }
}

核心规则:

• Node ID 是字符串（"1" 不是 1）
• 连接语法: ["source_node_id", output_index]
• 输出索引对应 RETURN_TYPES 的顺序（0-based）

---

构建步骤

Step 1: 查询环境

# 查可用模型
ckpts = folder_paths.get_filename_list("checkpoints")
print("Checkpoints:", ckpts)

# 查节点参数（第一次用某节点前必查）
info = nodes.NODE_CLASS_MAPPINGS["KSampler"].INPUT_TYPES()
print(info)

Step 2: 逐步构建 Workflow

import random

ckpt_name = ckpts[0]  # 选第一个模型
seed = random.randint(0, 2**63)

wf = {}
wf["1"] = {"class_type": "CheckpointLoaderSimple", "inputs": {"ckpt_name": ckpt_name}}
wf["2"] = {"class_type": "CLIPTextEncode", "inputs": {"text": "a beautiful sunset", "clip": ["1", 1]}}
wf["3"] = {"class_type": "CLIPTextEncode", "inputs": {"text": "ugly, blurry", "clip": ["1", 1]}}
wf["4"] = {"class_type": "EmptyLatentImage", "inputs": {"width": 1024, "height": 1024, "batch_size": 1}}
wf["5"] = {"class_type": "KSampler", "inputs": {
    "model": ["1", 0], "positive": ["2", 0], "negative": ["3", 0],
    "latent_image": ["4", 0], "seed": seed, "steps": 20,
    "cfg": 7.0, "sampler_name": "euler", "scheduler": "normal", "denoise": 1.0
}}
wf["6"] = {"class_type": "VAEDecode", "inputs": {"samples": ["5", 0], "vae": ["1", 2]}}
wf["7"] = {"class_type": "SaveImage", "inputs": {"images": ["6", 0], "filename_prefix": "ComfyUI"}}

Step 3: 提交并获取结果

result = submit_workflow(wf)
print(f"Prompt ID: {result['prompt_id']}")
print(f"Images: {len(result['images'])}")

# 展示输出
for img in result["images"]:
    img_bytes = client.get_image(img["filename"], img["subfolder"], img["type"])
    save_preview(img_bytes, f"output_{img['node_id']}")