Scale To Megapixels

Documentation

• Class name: ImageScaleToMegapixels
• Category: Art Venture/Utils
• Output node: False

The node is designed to adjust the size of an image to meet a specified total number of megapixels, either by upscaling or downscaling. It incorporates various upscaling methods to achieve the desired image quality and size, ensuring the final image adheres to the specified megapixel count while maintaining aspect ratio.

Input types

Required

• images
  • The input images to be resized. This parameter is crucial as it serves as the base for the scaling operation, determining the starting point for any upscaling or downscaling to achieve the target megapixel size.
  • Comfy dtype: IMAGE
  • Python dtype: torch.Tensor
• megapixels
  • Defines the target size of the image in megapixels, guiding the scaling process to either increase or decrease the image's dimensions to meet this specification.
  • Comfy dtype: FLOAT
  • Python dtype: float

Optional

• upscale_model_opt
  • Optionally specifies the upscale model to be used for upscaling the image, allowing for advanced customization of the image quality and appearance post-scaling. This parameter influences the visual outcome of the resized image, especially when upscale is required beyond a certain threshold.
  • Comfy dtype: UPSCALE_MODEL
  • Python dtype: str

Output types

• image
  • Comfy dtype: IMAGE
  • The resized image, adjusted to meet the specified megapixel count. This output is the direct result of the scaling operation, reflecting the changes in size and potentially quality based on the chosen upscaling method.
  • Python dtype: torch.Tensor

Usage tips

• Infra type: GPU
• Common nodes: unknown

Source code

class UtilImageScaleToTotalPixels(UtilImageScaleDownBy, ImageUpscaleWithModel):
    @classmethod
    def INPUT_TYPES(s):
        return {
            "required": {
                "images": ("IMAGE",),
                "megapixels": ("FLOAT", {"default": 1, "min": 0.1, "max": 100, "step": 0.05}),
            },
            "optional": {
                "upscale_model_opt": ("UPSCALE_MODEL",),
            },
        }

    RETURN_TYPES = ("IMAGE",)
    CATEGORY = "Art Venture/Utils"
    FUNCTION = "image_scale_down_to_total_pixels"

    def image_scale_up_by(self, images: torch.Tensor, scale_by, upscale_model_opt):
        width = round(images.shape[2] * scale_by)
        height = round(images.shape[1] * scale_by)

        if scale_by < 1.2 or upscale_model_opt is None:
            s = images.movedim(-1, 1)
            s = comfy.utils.common_upscale(s, width, height, "bicubic", "disabled")
            s = s.movedim(1, -1)
            return (s,)
        else:
            s = self.upscale(upscale_model_opt, images)[0]
            return self.image_scale_down(s, width, height, "center")

    def image_scale_down_to_total_pixels(self, images, megapixels, upscale_model_opt=None):
        width = images.shape[2]
        height = images.shape[1]
        scale_by = np.sqrt((megapixels * 1024 * 1024) / (width * height))

        if scale_by <= 1.0:
            return self.image_scale_down_by(images, scale_by)
        else:
            return self.image_scale_up_by(images, scale_by, upscale_model_opt)