Latent Scale by ratio

Documentation

• Class name: DF_Latent_Scale_by_ratio
• Category: Derfuu_Nodes/Modded nodes/Latent
• Output node: False

This node is designed to scale latent representations of images by a specified ratio, adjusting their dimensions while maintaining the integrity of the latent space. It supports various scaling and cropping methods to fine-tune the output.

Input types

Required

• latent
  • The latent representation of an image to be scaled. It is crucial for determining the base dimensions that will be modified according to the scaling ratio.
  • Comfy dtype: LATENT
  • Python dtype: Dict[str, Any]
• modifier
  • A scaling factor that determines how much the latent dimensions should be increased or decreased.
  • Comfy dtype: FLOAT
  • Python dtype: float
• scale_method
  • Specifies the method used for scaling the latent image, such as nearest-exact, bilinear, etc., affecting the quality and characteristics of the output.
  • Comfy dtype: COMBO[STRING]
  • Python dtype: str
• crop
  • Defines the cropping method to be applied after scaling, allowing for adjustments in the latent's aspect ratio or size.
  • Comfy dtype: COMBO[STRING]
  • Python dtype: str

Output types

• latent
  • Comfy dtype: LATENT
  • The scaled latent representation, adjusted according to the specified modifier, scale method, and cropping preference.
  • Python dtype: Dict[str, Any]

Usage tips

• Infra type: CPU
• Common nodes: unknown

Source code

class LatentScale_Ratio:
    scale_methods = scale_methods
    crop_methods = ["disabled", "center"]

    def __init__(self):
        pass

    @classmethod
    def INPUT_TYPES(cls):
        return {
            "required": {
                "latent": Field.latent(),
                "modifier": Field.float(min=0),
                "scale_method": Field.combo(cls.scale_methods),
                "crop": Field.combo(cls.crop_methods),
            }
        }

    RETURN_TYPES = ("LATENT",)
    FUNCTION = "scale"
    CATEGORY = TREE_LATENTS

    def scale(self, latent, scale_method, crop, modifier):

        size = get_latent_size(latent, True)

        lat_width = size[0] * modifier
        lat_width = int(lat_width + lat_width % 8)

        lat_height = size[1] * modifier
        lat_height = int(lat_height + lat_height % 8)

        cls = latent.copy()
        cls["samples"] = common_upscale(latent["samples"], lat_width, lat_height, scale_method, crop)
        return (cls,)