【修复】角色被格线切断(脚、手杖、飘起的斗篷被削掉) 新增 expand_beyond_cell(默认开启):格子只用来判定「这是哪个方向」, 角色的实际范围由它自身的连通区域决定,按质心归属确保邻居不混入。 实测 8/8 方向的裁剪框边缘 alpha 从 1.00(内容顶到边界=被切断) 降到 0.00,S 方向高度 326→356、E 方向宽度 150→188 把缺的部分找了回来。 代价是需要两遍扫描(先求全序列并集框再提取),耗时 4.7s→14.9s。 【规则】每个参数都必须有中文 tooltip,作为以后的统一约定 全仓库 26 个节点 169 个参数,此前缺 115 个,现已 100% 覆盖。 tooltip 写「怎么调」而不只是「是什么」:给取值区间的实际影响、 推荐值与踩坑提示(如 OpenAI/ZenMux 的地址栏不能带 :// , 素材拆分节点用于动画序列时顺序会漂移等)。 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
199 lines
7.7 KiB
Python
199 lines
7.7 KiB
Python
import torch
|
||
import numpy as np
|
||
import requests
|
||
import json
|
||
import base64
|
||
import io
|
||
from PIL import Image
|
||
import random
|
||
|
||
class QwenEditNode:
|
||
"""
|
||
使用阿里云千问编辑模型API进行图像生成的节点
|
||
"""
|
||
|
||
@classmethod
|
||
def INPUT_TYPES(cls):
|
||
return {
|
||
"required": {
|
||
"image1": ("IMAGE", {
|
||
"tooltip": "参考图 1(主参考)。控制模式决定它以什么方式影响生成,\n"
|
||
"例如 reference 取整体风格、pose 取人物姿态。"
|
||
}),
|
||
"image2": ("IMAGE", {
|
||
"tooltip": "参考图 2。多张参考会一并提交,用于综合约束生成结果。"
|
||
}),
|
||
"image3": ("IMAGE", {
|
||
"tooltip": "参考图 3。"
|
||
}),
|
||
"image4": ("IMAGE", {
|
||
"tooltip": "参考图 4。"
|
||
}),
|
||
"api_key": ("STRING", {
|
||
"default": "",
|
||
"multiline": False,
|
||
"tooltip": "阿里云百炼(DashScope)的 API Key。\n"
|
||
"⚠ 工作流会连同此值一起保存,分享 json 前记得清空。"
|
||
}),
|
||
"base_url": ("STRING", {
|
||
"default": "https://dashscope.aliyuncs.com/api/v1/services/aigc/text2image-generation/generation",
|
||
"multiline": False,
|
||
"tooltip": "接口地址,一般不用改。\n"
|
||
"走代理或私有网关时替换成对应地址。"
|
||
}),
|
||
"seed": ("INT", {
|
||
"default": -1,
|
||
"min": -1,
|
||
"max": 2147483647,
|
||
"tooltip": "随机种子。-1 表示每次随机;\n"
|
||
"填固定值可复现同一结果,便于对比参数改动的影响。"
|
||
}),
|
||
"control_mode": (["reference", "sketch", "scribble", "pose", "canny", "depth", "hed", "mlsd", "normal", "seg"], {
|
||
"default": "reference",
|
||
"tooltip": "参考图的约束方式:\n"
|
||
"reference 整体风格参考(最常用)\n"
|
||
"pose 取人物姿态 / depth 取空间深度 / seg 取区域分割\n"
|
||
"canny·hed·mlsd·scribble·sketch 取不同粗细的线稿\n"
|
||
"normal 取法线朝向。约束越强,构图越贴近参考图。"
|
||
}),
|
||
"width": ("INT", {
|
||
"default": 1024,
|
||
"min": 512,
|
||
"max": 2048,
|
||
"step": 8,
|
||
"tooltip": "输出宽度。与高度共同决定画幅比例,\n"
|
||
"比例与参考图差异过大时构图容易变形。"
|
||
}),
|
||
"height": ("INT", {
|
||
"default": 1024,
|
||
"min": 512,
|
||
"max": 2048,
|
||
"step": 8,
|
||
"tooltip": "输出高度。"
|
||
}),
|
||
},
|
||
}
|
||
|
||
RETURN_TYPES = ("IMAGE",)
|
||
FUNCTION = "generate_image"
|
||
CATEGORY = "Rui-Node🐶/AI模型🤖"
|
||
|
||
def generate_image(self, image1, image2, image3, image4, api_key, base_url, seed, control_mode, width, height):
|
||
"""
|
||
使用千问编辑模型API生成图像
|
||
|
||
参数:
|
||
image1-4: 输入图像张量 (B, H, W, C) 格式
|
||
api_key: 阿里云API密钥
|
||
base_url: API基础URL
|
||
seed: 随机种子值,-1表示随机生成
|
||
control_mode: 控制模式
|
||
width: 输出图像宽度
|
||
height: 输出图像高度
|
||
|
||
返回:
|
||
生成的图像张量
|
||
"""
|
||
# 如果seed为-1,则随机生成种子
|
||
if seed == -1:
|
||
seed = random.randint(0, 2147483647)
|
||
|
||
# 准备图像数据
|
||
images = [image1, image2, image3, image4]
|
||
image_data = []
|
||
|
||
for i, img in enumerate(images):
|
||
if img is None or img.shape[0] == 0:
|
||
continue
|
||
|
||
# 取批次中的第一张图像
|
||
img_np = img[0].cpu().numpy()
|
||
|
||
# 确保值范围在 0-1 之间
|
||
img_np = np.clip(img_np, 0, 1)
|
||
|
||
# 转换为 PIL 图像 (值范围 0-255)
|
||
img_pil = Image.fromarray((img_np * 255).astype(np.uint8), 'RGB')
|
||
|
||
# 转换为base64编码
|
||
buffered = io.BytesIO()
|
||
img_pil.save(buffered, format="PNG")
|
||
img_base64 = base64.b64encode(buffered.getvalue()).decode("utf-8")
|
||
|
||
# 添加到图像数据列表
|
||
image_data.append({
|
||
"image": img_base64,
|
||
"control_type": control_mode
|
||
})
|
||
|
||
# 准备API请求
|
||
headers = {
|
||
"Content-Type": "application/json",
|
||
"Authorization": f"Bearer {api_key}"
|
||
}
|
||
|
||
payload = {
|
||
"model": "qwen-vl-plus",
|
||
"input": {
|
||
"images": image_data
|
||
},
|
||
"parameters": {
|
||
"seed": seed,
|
||
"width": width,
|
||
"height": height
|
||
}
|
||
}
|
||
|
||
try:
|
||
# 发送API请求
|
||
response = requests.post(base_url, headers=headers, data=json.dumps(payload))
|
||
response.raise_for_status()
|
||
|
||
# 解析响应
|
||
result = response.json()
|
||
|
||
# 检查是否成功
|
||
if "output" in result and "images" in result["output"] and len(result["output"]["images"]) > 0:
|
||
# 获取生成的图像
|
||
generated_image_base64 = result["output"]["images"][0]
|
||
|
||
# 解码base64图像
|
||
image_bytes = base64.b64decode(generated_image_base64)
|
||
img_pil = Image.open(io.BytesIO(image_bytes))
|
||
|
||
# 确保图像为 RGB 模式
|
||
if img_pil.mode != 'RGB':
|
||
img_pil = img_pil.convert('RGB')
|
||
|
||
# 转换为 NumPy 数组
|
||
img_np = np.array(img_pil).astype(np.float32) / 255.0
|
||
|
||
# 转换为 PyTorch 张量并添加批次维度
|
||
img_tensor = torch.from_numpy(img_np).unsqueeze(0)
|
||
|
||
return (img_tensor,)
|
||
else:
|
||
# 如果响应中没有图像,返回错误信息
|
||
print(f"API响应中没有图像: {result}")
|
||
# 创建一个默认的黑色图像作为返回值
|
||
default_img = np.zeros((height, width, 3), dtype=np.float32)
|
||
default_tensor = torch.from_numpy(default_img).unsqueeze(0)
|
||
return (default_tensor,)
|
||
|
||
except Exception as e:
|
||
# 如果API请求失败,打印错误信息并返回默认图像
|
||
print(f"API请求失败: {str(e)}")
|
||
# 创建一个默认的黑色图像作为返回值
|
||
default_img = np.zeros((height, width, 3), dtype=np.float32)
|
||
default_tensor = torch.from_numpy(default_img).unsqueeze(0)
|
||
return (default_tensor,)
|
||
|
||
# 节点映射字典,用于 ComfyUI 注册节点
|
||
NODE_CLASS_MAPPINGS = {
|
||
"QwenEditImageGeneration": QwenEditNode
|
||
}
|
||
|
||
# 节点显示名称映射,用于在 UI 中显示友好名称
|
||
NODE_DISPLAY_NAME_MAPPINGS = {
|
||
"QwenEditImageGeneration": "千问编辑图像生成 / Qwen Edit Image Generation"
|
||
} |