ComfyUI_TravelSuite

ComfyUI_TravelSuite
★ 16

潜在空间混合模式过渡生成节点扩展
ComfyUI自定义节点组,提供在潜在空间中多种旅行与混合技术,支持lerp/slerp/add等模式,便于生成平滑过渡与多样化变体。
💡 用于在潜在空间中生成多步平滑过渡或混合变体。
🍴 9 Forks💻 Python🔄 2024-05-22
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/c73fe210bae7
📦 requirements.txt
torch
numpy
Pillow
scipy
📄 README

ComfyUI_TravelSuite

ComfyUI custom nodes to apply various latent travel techniques.

LatentTravel Node

Travel between different latent spaces using a range of blend and travel modes.

___Node Inputs___

  • A, B: Latent variables needed for the process.
  • steps: Integer representing the number of steps. This includes the two input latents A and B which will always be the first and last latents. The default value is 5. It should be within the range [0,10000].
  • factor: A floating-point value with a default of 0.5. This value depends on the travel_mode specified
  • blend_mode: Specifies the method used for latent blending: lerp”, “slerp”, “add”, “multiply”, “divide”, “subtract”, “overlay”, “hard_light”, “soft_light”, “screen”, “linear_dodge”, “difference”, “exclusion”, “random”
  • travel_mode: Determines the travel interval model:
  • – ‘linear’: linear interpolation
  • – ‘hinge’: takes factor as a cutpoint and splits the steps evenly between above and below the hinge
  • – ‘circle’: steps along the X-axis of a circle
  • – ‘norm’: steps along a normal distribution centered at 0.5. Factor is the gaussian scale factor.
  • – ‘quadratic’, ‘cubic’, ‘quartic’: steps along linear space raised to 2, 3, 4 powers.
  • – ‘geometric’: steps along geomspace from 0, 1
  • reflect_travel: Bool indicating whether to reflect travel mode values around the center.
  • vae: VAE (Variational Autoencoder) type, to decode generated latents to tensors.
  • output_images: Bool indicating whether to output images. Requires VAE input.
  • filepath: String for defining the output path for the files. Default path is ‘output/travel’.
  • prefix: String for prefixing the generated files. By default, this is ‘travel’.
  • write_images: Bool indicating whether to write output images. Requires VAE input.
  • ___Node Outputs___

  • LATENTS: Resulting travel latents. There will be steps latents, and the first and last will be the input A and B.
  • IMAGES: Tensor images, if output_images=True. Requires VAE.
  • FILEPATHS: List of filepaths produced if write_images=True.
  • Workflow Image