514 lines
22 KiB
Python
514 lines
22 KiB
Python
from .md import *
|
|
CATEGORY_TYPE = "🎈LAOGOU/Utils"
|
|
|
|
pb_id_cnt = time.time()
|
|
preview_bridge_image_id_map = {}
|
|
preview_bridge_image_name_map = {}
|
|
preview_bridge_cache = {}
|
|
preview_bridge_last_mask_cache = {}
|
|
|
|
def set_previewbridge_image(node_id, file, item):
|
|
global pb_id_cnt
|
|
|
|
if file in preview_bridge_image_name_map:
|
|
pb_id = preview_bridge_image_name_map[node_id, file]
|
|
if pb_id.startswith(f"${node_id}"):
|
|
return pb_id
|
|
|
|
pb_id = f"${node_id}-{pb_id_cnt}"
|
|
|
|
preview_bridge_image_id_map[pb_id] = (file, item)
|
|
preview_bridge_image_name_map[node_id, file] = (pb_id, item)
|
|
|
|
pb_id_cnt += 1
|
|
|
|
return pb_id
|
|
|
|
|
|
class CachePreviewBridge:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {"required": {
|
|
"image": ("STRING", {"default": ""}),
|
|
"use_cache": ("BOOLEAN", {"default": True, "label_on": "Cache", "label_off": "Input"}),
|
|
},
|
|
"optional": {
|
|
"images": ("IMAGE",),
|
|
},
|
|
"hidden": {"unique_id": "UNIQUE_ID", "extra_pnginfo": "EXTRA_PNGINFO"},
|
|
}
|
|
|
|
RETURN_TYPES = ("IMAGE", "MASK")
|
|
FUNCTION = "doit"
|
|
OUTPUT_NODE = True
|
|
CATEGORY = CATEGORY_TYPE
|
|
|
|
def __init__(self):
|
|
super().__init__()
|
|
self.output_dir = folder_paths.get_temp_directory()
|
|
self.type = "temp"
|
|
self.prev_hash = None
|
|
@staticmethod
|
|
def load_image(pb_id):
|
|
clipspace_dir = os.path.join(folder_paths.get_input_directory(), "clipspace")
|
|
files = [f for f in os.listdir(clipspace_dir)
|
|
if f.lower().startswith('clipspace') and f.lower().endswith('.png')]
|
|
|
|
# 初始化默认值
|
|
image = torch.zeros((1, 512, 512, 3), dtype=torch.float32, device="cpu")
|
|
mask = torch.zeros((512, 512), dtype=torch.float32, device="cpu")
|
|
ui_item = {
|
|
"filename": 'empty.png',
|
|
"subfolder": '',
|
|
"type": 'temp'
|
|
}
|
|
|
|
if files:
|
|
latest_file = max(files, key=lambda f: os.path.getmtime(os.path.join(clipspace_dir, f)))
|
|
latest_path = os.path.join(clipspace_dir, latest_file)
|
|
latest_mtime = os.path.getmtime(latest_path)
|
|
|
|
current_path = None
|
|
if pb_id in preview_bridge_image_id_map:
|
|
current_path, ui_item = preview_bridge_image_id_map[pb_id]
|
|
current_mtime = os.path.getmtime(current_path) if os.path.exists(current_path) else 0
|
|
|
|
if current_path is None or latest_mtime > current_mtime:
|
|
preview_dir = os.path.join(folder_paths.get_temp_directory(), 'PreviewBridge')
|
|
os.makedirs(preview_dir, exist_ok=True)
|
|
|
|
new_filename = f"PB-{os.path.splitext(latest_file)[0]}.png"
|
|
new_path = os.path.join(preview_dir, new_filename)
|
|
|
|
# 检查是否存在同名文件
|
|
if os.path.exists(new_path):
|
|
if pb_id in preview_bridge_image_id_map:
|
|
current_path, ui_item = preview_bridge_image_id_map[pb_id]
|
|
if os.path.exists(current_path):
|
|
new_path = current_path
|
|
else:
|
|
import shutil
|
|
shutil.copy2(latest_path, new_path)
|
|
ui_item = {
|
|
"filename": new_filename,
|
|
"subfolder": 'PreviewBridge',
|
|
"type": 'temp'
|
|
}
|
|
preview_bridge_image_id_map[pb_id] = (new_path, ui_item)
|
|
else:
|
|
import shutil
|
|
shutil.copy2(latest_path, new_path)
|
|
ui_item = {
|
|
"filename": new_filename,
|
|
"subfolder": 'PreviewBridge',
|
|
"type": 'temp'
|
|
}
|
|
preview_bridge_image_id_map[pb_id] = (new_path, ui_item)
|
|
|
|
current_path = new_path
|
|
|
|
try:
|
|
i = Image.open(current_path)
|
|
i = ImageOps.exif_transpose(i)
|
|
image = i.convert("RGB")
|
|
image = np.array(image).astype(np.float32) / 255.0
|
|
image = torch.from_numpy(image)[None,]
|
|
|
|
if 'A' in i.getbands():
|
|
mask = np.array(i.getchannel('A')).astype(np.float32) / 255.0
|
|
mask = 1. - torch.from_numpy(mask)
|
|
else:
|
|
mask = torch.zeros((image.shape[1], image.shape[2]), dtype=torch.float32, device="cpu")
|
|
except Exception as e:
|
|
print(f"Error loading image: {e}")
|
|
# 出错时使用默认值
|
|
|
|
final_mask = mask.unsqueeze(0) if len(mask.shape) == 2 else mask
|
|
return image, final_mask, ui_item
|
|
|
|
def doit(self, image, use_cache, unique_id, images=None, extra_pnginfo=None):
|
|
# 检查是否有新的剪贴板图片需要更新缓存
|
|
if images is None and image and image not in preview_bridge_image_id_map:
|
|
print("Component value changed, updating cache")
|
|
# 生成新的pb_id并更新缓存
|
|
pb_id = f"${unique_id}-{time.time()}"
|
|
pixels, mask, path_item = CachePreviewBridge.load_image(pb_id)
|
|
if path_item["type"] != "temp" or path_item["filename"] != "empty.png":
|
|
return {
|
|
"ui": {"images": [path_item]},
|
|
"result": (pixels, mask),
|
|
}
|
|
|
|
# 原有的判断逻辑
|
|
if images is None and not image:
|
|
empty_image = torch.zeros((1, 512, 512, 3), dtype=torch.float32, device="cpu")
|
|
empty_mask = torch.zeros((1, 512, 512), dtype=torch.float32, device="cpu")
|
|
return {
|
|
"ui": {"images": []},
|
|
"result": (empty_image, empty_mask),
|
|
}
|
|
|
|
if use_cache:
|
|
if not image:
|
|
|
|
empty_image = torch.zeros((1, 512, 512, 3), dtype=torch.float32, device="cpu")
|
|
empty_mask = torch.zeros((1, 512, 512), dtype=torch.float32, device="cpu")
|
|
return {
|
|
"ui": {"images": []},
|
|
"result": (empty_image, empty_mask),
|
|
}
|
|
|
|
if image.startswith('$'):
|
|
node_id = image.split('-')[0][1:]
|
|
related_pb_ids = [pb_id for pb_id in preview_bridge_image_id_map.keys()
|
|
if pb_id.startswith(f"${node_id}-")]
|
|
|
|
if related_pb_ids:
|
|
latest_pb_id = max(related_pb_ids, key=lambda x: float(x.split('-')[1]))
|
|
image = latest_pb_id
|
|
pixels, mask, path_item = CachePreviewBridge.load_image(image)
|
|
return {
|
|
"ui": {"images": [path_item]},
|
|
"result": (pixels, mask),
|
|
}
|
|
|
|
# 非缓存模式或需要创建新缓存
|
|
if images is not None:
|
|
mask = torch.zeros((1, images.shape[1], images.shape[2]), dtype=torch.float32, device="cpu")
|
|
res = PreviewImage().save_images(
|
|
images,
|
|
filename_prefix=f"PreviewBridge/PB-{unique_id}-",
|
|
extra_pnginfo=extra_pnginfo
|
|
)
|
|
|
|
image2 = res['ui']['images']
|
|
pixels = images
|
|
|
|
path = os.path.join(folder_paths.get_temp_directory(), 'PreviewBridge', image2[0]['filename'])
|
|
pb_id = set_previewbridge_image(unique_id, path, image2[0])
|
|
|
|
preview_bridge_cache[unique_id] = (images, image2)
|
|
|
|
return {
|
|
"ui": {"images": image2},
|
|
"result": (pixels, mask),
|
|
}
|
|
|
|
# 如果走到这里,说明既没有有效的缓存也没有新的图像输入
|
|
empty_image = torch.zeros((1, 512, 512, 3), dtype=torch.float32, device="cpu")
|
|
empty_mask = torch.zeros((1, 512, 512), dtype=torch.float32, device="cpu")
|
|
return {
|
|
"ui": {"images": []},
|
|
"result": (empty_image, empty_mask),
|
|
}
|
|
|
|
class LG_Noise:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {
|
|
"required": {
|
|
"type": (["fade", "dissolve", "gaussian"], ),
|
|
"opacity": ("FLOAT", { "default": 1.0, "min": 0, "max": 1, "step": 0.01 }),
|
|
"strength": ("INT", { "default": 1, "min": 1, "max": 32, "step": 1 }),
|
|
"density": ("FLOAT", { "default": 1.0, "min": 0, "max": 1, "step": 0.05 }),
|
|
"sharpen": ("INT", { "default": 0, "min": -32, "max": 32, "step": 1 }),
|
|
"brightness": ("FLOAT", { "default": 1.0, "min": 0, "max": 3, "step": 0.05 }),
|
|
"random_color": ("BOOLEAN", {"default": True}),
|
|
"color": ("COLOR", {"default": "#808080"}),
|
|
"seed": ("INT", {"default": -1, "min": -1, "max": 0x7fffffff}),
|
|
},
|
|
"optional": {
|
|
"image_optional": ("IMAGE",),
|
|
"mask_optional": ("MASK",),
|
|
}
|
|
}
|
|
RETURN_TYPES = ("IMAGE",)
|
|
FUNCTION = "make_noise"
|
|
CATEGORY = CATEGORY_TYPE
|
|
def hex_to_rgb(self, hex_color):
|
|
hex_color = hex_color.lstrip('#')
|
|
r = int(hex_color[0:2], 16) / 255.0
|
|
g = int(hex_color[2:4], 16) / 255.0
|
|
b = int(hex_color[4:6], 16) / 255.0
|
|
return torch.tensor([r, g, b])
|
|
def make_noise(self, type, opacity, density, strength, sharpen, random_color, color, brightness, seed,
|
|
image_optional=None, mask_optional=None):
|
|
if image_optional is None:
|
|
image = torch.zeros([1, 1, 1, 3])
|
|
else:
|
|
image = image_optional
|
|
h, w = image.shape[1:3]
|
|
if h == 1 and w == 1:
|
|
image = image.repeat(1, 512, 512, 1)
|
|
h, w = 512, 512
|
|
if mask_optional is not None:
|
|
mask = mask_optional.unsqueeze(-1).repeat(1, 1, 1, 3)
|
|
else:
|
|
mask = torch.ones((1, h, w, 1), device=image.device).repeat(1, 1, 1, 3)
|
|
if seed == -1:
|
|
seed = torch.randint(0, 0x7fffffff, (1,)).item()
|
|
torch.manual_seed(seed)
|
|
print(f"[LG_Noise] Using seed: {seed}")
|
|
color_rgb = self.hex_to_rgb(color).to(image.device)
|
|
def generate_noise(size_h, size_w, is_gaussian=False):
|
|
"""统一的噪声生成函数"""
|
|
if random_color:
|
|
if is_gaussian:
|
|
noise = torch.randn(1, size_h, size_w, 3, device=image.device)
|
|
return torch.clamp(noise * 0.5 + 0.5, 0, 1)
|
|
else:
|
|
return torch.rand(1, size_h, size_w, 3, device=image.device)
|
|
else:
|
|
return torch.ones(1, size_h, size_w, 3, device=image.device)
|
|
def generate_density_mask(size_h, size_w):
|
|
"""统一的密度遮罩生成函数"""
|
|
return (torch.rand(1, size_h, size_w, 1, device=image.device) < density).float()
|
|
if strength > 1:
|
|
small_h, small_w = h // strength, w // strength
|
|
density_mask = generate_density_mask(small_h, small_w)
|
|
noise = generate_noise(small_h, small_w) * density_mask
|
|
noise = torch.nn.functional.interpolate(
|
|
noise.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
if type != "fade":
|
|
density_mask = torch.nn.functional.interpolate(
|
|
density_mask.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
else:
|
|
density_mask = generate_density_mask(h, w)
|
|
noise = generate_noise(h, w) * density_mask
|
|
colored_noise = noise * color_rgb.view(1, 1, 1, 3) * brightness
|
|
if type == "fade":
|
|
result = image * (1 - opacity) + colored_noise * opacity
|
|
elif type == "dissolve":
|
|
dissolve_mask = (torch.rand(1, h//strength if strength > 1 else h,
|
|
w//strength if strength > 1 else w, 1,
|
|
device=image.device) < opacity).float()
|
|
density_mask = (torch.rand(1, h//strength if strength > 1 else h,
|
|
w//strength if strength > 1 else w, 1,
|
|
device=image.device) < density).float()
|
|
dissolve_mask = dissolve_mask * density_mask
|
|
if strength > 1:
|
|
dissolve_mask = torch.nn.functional.interpolate(
|
|
dissolve_mask.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
dissolve_mask = dissolve_mask.repeat(1, 1, 1, 3)
|
|
if random_color:
|
|
noise = torch.rand(1, h, w, 3, device=image.device)
|
|
else:
|
|
noise = torch.ones(1, h, w, 3, device=image.device)
|
|
noise = noise * dissolve_mask
|
|
colored_noise = noise * color_rgb.view(1, 1, 1, 3) * brightness
|
|
result = image * (1-dissolve_mask) + colored_noise * dissolve_mask
|
|
elif type == "gaussian":
|
|
noise = generate_noise(h//strength if strength > 1 else h,
|
|
w//strength if strength > 1 else w,
|
|
is_gaussian=True)
|
|
if strength > 1:
|
|
noise = torch.nn.functional.interpolate(
|
|
noise.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
noise = (noise - 0.5) * opacity * 2
|
|
noise = noise * density_mask
|
|
colored_noise = noise * color_rgb.view(1, 1, 1, 3) * brightness
|
|
result = torch.clamp(image + colored_noise, 0, 1)
|
|
result = torch.clamp(result, 0, 1)
|
|
if sharpen != 0:
|
|
kernel_size = abs(sharpen) * 2 + 1
|
|
noise_part = result * mask - image * mask
|
|
if sharpen < 0:
|
|
blurred_noise = T.functional.gaussian_blur(
|
|
noise_part.permute([0,3,1,2]),
|
|
kernel_size
|
|
).permute([0,2,3,1])
|
|
result = image + blurred_noise
|
|
else:
|
|
blurred = T.functional.gaussian_blur(
|
|
noise_part.permute([0,3,1,2]),
|
|
kernel_size
|
|
).permute([0,2,3,1])
|
|
sharpened_noise = noise_part + (noise_part - blurred) * (sharpen / 8)
|
|
result = image + torch.clamp(sharpened_noise, 0, 1)
|
|
result = torch.clamp(result, 0, 1)
|
|
result = image * (1 - mask) + result * mask
|
|
return (result,)
|
|
|
|
class LG_Noise:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {
|
|
"required": {
|
|
"type": (["fade", "dissolve", "gaussian"], ),
|
|
"opacity": ("FLOAT", { "default": 1.0, "min": 0, "max": 1, "step": 0.01 }),
|
|
"strength": ("INT", { "default": 1, "min": 1, "max": 32, "step": 1 }),
|
|
"density": ("FLOAT", { "default": 1.0, "min": 0, "max": 1, "step": 0.05 }),
|
|
"sharpen": ("INT", { "default": 0, "min": -32, "max": 32, "step": 1 }),
|
|
"brightness": ("FLOAT", { "default": 1.0, "min": 0, "max": 3, "step": 0.05 }),
|
|
"random_color": ("BOOLEAN", {"default": True}),
|
|
"color": ("COLOR", {"default": "#808080"}),
|
|
"seed": ("INT", {"default": -1, "min": -1, "max": 0x7fffffff}),
|
|
},
|
|
"optional": {
|
|
"image_optional": ("IMAGE",),
|
|
"mask_optional": ("MASK",),
|
|
}
|
|
}
|
|
RETURN_TYPES = ("IMAGE",)
|
|
FUNCTION = "make_noise"
|
|
CATEGORY = CATEGORY_TYPE
|
|
def hex_to_rgb(self, hex_color):
|
|
hex_color = hex_color.lstrip('#')
|
|
r = int(hex_color[0:2], 16) / 255.0
|
|
g = int(hex_color[2:4], 16) / 255.0
|
|
b = int(hex_color[4:6], 16) / 255.0
|
|
return torch.tensor([r, g, b])
|
|
def make_noise(self, type, opacity, density, strength, sharpen, random_color, color, brightness, seed,
|
|
image_optional=None, mask_optional=None):
|
|
if image_optional is None:
|
|
image = torch.zeros([1, 1, 1, 3])
|
|
else:
|
|
image = image_optional
|
|
h, w = image.shape[1:3]
|
|
if h == 1 and w == 1:
|
|
image = image.repeat(1, 512, 512, 1)
|
|
h, w = 512, 512
|
|
if mask_optional is not None:
|
|
mask = mask_optional.unsqueeze(-1).repeat(1, 1, 1, 3)
|
|
else:
|
|
mask = torch.ones((1, h, w, 1), device=image.device).repeat(1, 1, 1, 3)
|
|
if seed == -1:
|
|
seed = torch.randint(0, 0x7fffffff, (1,)).item()
|
|
torch.manual_seed(seed)
|
|
print(f"[LG_Noise] Using seed: {seed}")
|
|
color_rgb = self.hex_to_rgb(color).to(image.device)
|
|
def generate_noise(size_h, size_w, is_gaussian=False):
|
|
"""统一的噪声生成函数"""
|
|
if random_color:
|
|
if is_gaussian:
|
|
noise = torch.randn(1, size_h, size_w, 3, device=image.device)
|
|
return torch.clamp(noise * 0.5 + 0.5, 0, 1)
|
|
else:
|
|
return torch.rand(1, size_h, size_w, 3, device=image.device)
|
|
else:
|
|
return torch.ones(1, size_h, size_w, 3, device=image.device)
|
|
def generate_density_mask(size_h, size_w):
|
|
"""统一的密度遮罩生成函数"""
|
|
return (torch.rand(1, size_h, size_w, 1, device=image.device) < density).float()
|
|
if strength > 1:
|
|
small_h, small_w = h // strength, w // strength
|
|
density_mask = generate_density_mask(small_h, small_w)
|
|
noise = generate_noise(small_h, small_w) * density_mask
|
|
noise = torch.nn.functional.interpolate(
|
|
noise.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
if type != "fade":
|
|
density_mask = torch.nn.functional.interpolate(
|
|
density_mask.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
else:
|
|
density_mask = generate_density_mask(h, w)
|
|
noise = generate_noise(h, w) * density_mask
|
|
colored_noise = noise * color_rgb.view(1, 1, 1, 3) * brightness
|
|
if type == "fade":
|
|
result = image * (1 - opacity) + colored_noise * opacity
|
|
elif type == "dissolve":
|
|
dissolve_mask = (torch.rand(1, h//strength if strength > 1 else h,
|
|
w//strength if strength > 1 else w, 1,
|
|
device=image.device) < opacity).float()
|
|
density_mask = (torch.rand(1, h//strength if strength > 1 else h,
|
|
w//strength if strength > 1 else w, 1,
|
|
device=image.device) < density).float()
|
|
dissolve_mask = dissolve_mask * density_mask
|
|
if strength > 1:
|
|
dissolve_mask = torch.nn.functional.interpolate(
|
|
dissolve_mask.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
dissolve_mask = dissolve_mask.repeat(1, 1, 1, 3)
|
|
if random_color:
|
|
noise = torch.rand(1, h, w, 3, device=image.device)
|
|
else:
|
|
noise = torch.ones(1, h, w, 3, device=image.device)
|
|
noise = noise * dissolve_mask
|
|
colored_noise = noise * color_rgb.view(1, 1, 1, 3) * brightness
|
|
result = image * (1-dissolve_mask) + colored_noise * dissolve_mask
|
|
elif type == "gaussian":
|
|
noise = generate_noise(h//strength if strength > 1 else h,
|
|
w//strength if strength > 1 else w,
|
|
is_gaussian=True)
|
|
if strength > 1:
|
|
noise = torch.nn.functional.interpolate(
|
|
noise.permute(0, 3, 1, 2),
|
|
size=(h, w),
|
|
mode='nearest'
|
|
).permute(0, 2, 3, 1)
|
|
noise = (noise - 0.5) * opacity * 2
|
|
noise = noise * density_mask
|
|
colored_noise = noise * color_rgb.view(1, 1, 1, 3) * brightness
|
|
result = torch.clamp(image + colored_noise, 0, 1)
|
|
result = torch.clamp(result, 0, 1)
|
|
if sharpen != 0:
|
|
kernel_size = abs(sharpen) * 2 + 1
|
|
noise_part = result * mask - image * mask
|
|
if sharpen < 0:
|
|
blurred_noise = T.functional.gaussian_blur(
|
|
noise_part.permute([0,3,1,2]),
|
|
kernel_size
|
|
).permute([0,2,3,1])
|
|
result = image + blurred_noise
|
|
else:
|
|
blurred = T.functional.gaussian_blur(
|
|
noise_part.permute([0,3,1,2]),
|
|
kernel_size
|
|
).permute([0,2,3,1])
|
|
sharpened_noise = noise_part + (noise_part - blurred) * (sharpen / 8)
|
|
result = image + torch.clamp(sharpened_noise, 0, 1)
|
|
result = torch.clamp(result, 0, 1)
|
|
result = image * (1 - mask) + result * mask
|
|
return (result,)
|
|
|
|
|
|
|
|
WEIGHT_TYPES = ["linear", "ease in", "ease out", 'ease in-out', 'reverse in-out', 'weak input', 'weak output', 'weak middle', 'strong middle', 'style transfer', 'composition', 'strong style transfer', 'style and composition', 'style transfer precise', 'composition precise']
|
|
|
|
class IPAdapterWeightTypes:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {"required": {
|
|
"weight_type": (WEIGHT_TYPES, ),
|
|
}}
|
|
|
|
RETURN_TYPES = (AlwaysEqualProxy('*'),)
|
|
RETURN_NAMES = ("weight_type",)
|
|
FUNCTION = "get_weight_types"
|
|
CATEGORY = CATEGORY_TYPE
|
|
|
|
def get_weight_types(self, weight_type):
|
|
return (weight_type,)
|
|
|
|
NODE_CLASS_MAPPINGS = {
|
|
"CachePreviewBridge": CachePreviewBridge,
|
|
"LG_Noise": LG_Noise,
|
|
"IPAdapterWeightTypes": IPAdapterWeightTypes
|
|
}
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"CachePreviewBridge": "🎈LG_PreviewBridge",
|
|
"LG_Noise": "🎈LG_Noise",
|
|
"IPAdapterWeightTypes": "🎈IPAdapter权重类型"
|
|
}
|
|
|