Wan2.2-TI2V-5B-Diffusers/WanImageProcessor/block.py

from diffusers.modular_pipelines import (
    ModularPipelineBlocks,
    InputParam,
    OutputParam,
    ConfigSpec,
)

from diffusers.utils import load_image
from PIL import Image
from typing import Union, Tuple

# copied from https://github.com/Wan-Video/Wan2.2/blob/388807310646ed5f318a99f8e8d9ad28c5b65373/wan/utils/utils.py#L136
def best_output_size(w, h, dw, dh, expected_area):
    # float output size
    ratio = w / h
    ow = (expected_area * ratio)**0.5
    oh = expected_area / ow

    # process width first
    ow1 = int(ow // dw * dw)
    oh1 = int(expected_area / ow1 // dh * dh)
    assert ow1 % dw == 0 and oh1 % dh == 0 and ow1 * oh1 <= expected_area
    ratio1 = ow1 / oh1

    # process height first
    oh2 = int(oh // dh * dh)
    ow2 = int(expected_area / oh2 // dw * dw)
    assert oh2 % dh == 0 and ow2 % dw == 0 and ow2 * oh2 <= expected_area
    ratio2 = ow2 / oh2

    # compare ratios
    if max(ratio / ratio1, ratio1 / ratio) < max(ratio / ratio2,
                                                 ratio2 / ratio):
        return ow1, oh1
    else:
        return ow2, oh2

class Wan225BI2VImageProcessor(ModularPipelineBlocks):

    @property
    def description(self):
        return "default Image Processor for wan2.2 5b i2v, it resizes the image to the best output size and center-crop it"

    @property
    def inputs(self):
        return [
            InputParam(name="image", type_hint=Union[Image.Image, str], description= "the Image to process"),
            InputParam(name="max_area", type_hint=int, description= "the maximum area of the Image to process")
        ]
    
    @property
    def intermediate_outputs(self):
        return [
            OutputParam(name="processed_image", type_hint=Image.Image, description= "the processed Image"),
        ]
    
    @property
    def expected_configs(self):
        return [
            ConfigSpec(name="patch_size", default=(1, 2, 2)),
            ConfigSpec(name="vae_stride", default=(4, 16, 16)),
        ]
    
    def __call__(self, components, state):

        block_state = self.get_block_state(state)

        if isinstance(block_state.image, str):
            image = load_image(block_state.image).convert("RGB")
        elif isinstance(block_state.image, Image.Image):
            image = block_state.image
        else:
            raise ValueError(f"Invalid image type: {type(block_state.image)}; only support PIL Image or url string")
        
        ih, iw = image.height, image.width
        dh, dw = components.patch_size[1] * components.vae_stride[1], components.patch_size[2] * components.vae_stride[2]
        ow, oh = best_output_size(iw, ih, dw, dh, block_state.max_area)

        scale = max(ow / iw, oh / ih)
        resized_image = image.resize((round(iw * scale), round(ih * scale)), Image.LANCZOS)

        # center-crop
        x1 = (resized_image.width - ow) // 2
        y1 = (resized_image.height - oh) // 2
        cropped_image = resized_image.crop((x1, y1, x1 + ow, y1 + oh))
        assert cropped_image.width == ow and cropped_image.height == oh

        block_state.processed_image = cropped_image

        print(f" initial image size: {image.size}")
        print(f" processed image size: {cropped_image.size}")
        

        self.set_block_state(state, block_state)
        return components, state