Skip to content

Generation Resolution From Latent

Documentation

  • Class name: ImageGenResolutionFromLatent
  • Category: ControlNet Preprocessors
  • Output node: False

This node is designed to calculate the generation resolution for images based on the dimensions of a given latent representation. It extracts the height and width from the latent's shape and scales them to determine the appropriate resolution for image generation.

Input types

Required

  • latent
    • The latent representation from which the image generation resolution will be derived. The latent's shape is used to calculate the desired output dimensions by scaling its height and width.
    • Comfy dtype: LATENT
    • Python dtype: Dict[str, torch.Tensor]

Output types

  • IMAGE_GEN_WIDTH (INT)
    • Comfy dtype: INT
    • The calculated width for image generation, derived from the latent's dimensions and scaled appropriately.
    • Python dtype: int
  • IMAGE_GEN_HEIGHT (INT)
    • Comfy dtype: INT
    • The calculated height for image generation, derived from the latent's dimensions and scaled appropriately.
    • Python dtype: int

Usage tips

Source code

class ImageGenResolutionFromLatent:
    @classmethod
    def INPUT_TYPES(s):
        return {
            "required": { "latent": ("LATENT", ) }
        }

    RETURN_TYPES = ("INT", "INT")
    RETURN_NAMES = ("IMAGE_GEN_WIDTH (INT)", "IMAGE_GEN_HEIGHT (INT)")
    FUNCTION = "execute"

    CATEGORY = "ControlNet Preprocessors"

    def execute(self, latent):
        _, _, H, W = latent["samples"].shape
        return (W * 8, H * 8)