65 lines
2.1 KiB
Python
65 lines
2.1 KiB
Python
import torch
|
|
|
|
from .blend_modes.index import blend_functions
|
|
from .blend_modes_enum import BlendModes
|
|
|
|
class ImageBlender:
|
|
def __init__(self):
|
|
self.blend_functions = blend_functions
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"base_image": ("IMAGE",),
|
|
"blend_image": ("IMAGE",),
|
|
"strength": ("FLOAT", {
|
|
"default": 1,
|
|
"min": 0.0,
|
|
"max": 1.0,
|
|
"step": 0.01
|
|
}),
|
|
"blend_mode": (
|
|
[mode.value for mode in BlendModes],
|
|
{"default": BlendModes.MIX_NORMAL.value}
|
|
),
|
|
},
|
|
"optional": {
|
|
"mask": ("MASK",),
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("IMAGE",)
|
|
FUNCTION = "blend"
|
|
CATEGORY = "ImageBlender"
|
|
|
|
def blend(self, base_image: torch.Tensor, blend_image: torch.Tensor, strength: float, blend_mode: str, mask: torch.Tensor = None) -> tuple:
|
|
assert base_image.shape == blend_image.shape, "Base and blend images must have the same shape"
|
|
assert base_image.shape[-1] == 3, "Input images must have 3 channels (RGB)"
|
|
|
|
blend_function = self.blend_functions.get(BlendModes(blend_mode), lambda x, y: x)
|
|
result = blend_function(base_image, blend_image)
|
|
|
|
if mask is not None:
|
|
# Ensure mask has the same number of channels as the images
|
|
if mask.dim() == 3:
|
|
mask = mask.unsqueeze(-1).expand(-1, -1, -1, base_image.shape[-1])
|
|
|
|
if mask.size() != base_image.size():
|
|
print(f"WARN: Mask size {mask.size()} is different from image size {base_image.size()}, mask is ignored")
|
|
else:
|
|
result = result * mask + base_image * (1 - mask)
|
|
|
|
# Apply opacity
|
|
result = result * strength + base_image * (1 - strength)
|
|
# Normalize the result
|
|
result = torch.clamp(result, 0, 1)
|
|
return (result,)
|
|
|
|
NODE_CLASS_MAPPINGS = {
|
|
"ImageBlender": ImageBlender
|
|
}
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"ImageBlender": "ImageBlender"
|
|
} |