latentblending / parameters.md
Johannes Stelzer
new latent blending with diffusers, xl, ...
940cc9a

Gradio Parameters

depth_strength

  • Determines when the blending process will begin in terms of diffusion steps.
  • A value close to zero results in more creative and intricate outcomes, but may also introduce additional objects and motion.
  • A value closer to one indicates a simpler alpha blending.

branch1_crossfeed_power

  • Controls the level of cross-feeding between the first and last image branch. This allows to preserve structures from the first image.
  • A value of 0.0 disables crossfeeding.
  • A value of 1.0 fully copies the latents from the first branch to the last.

branch1_crossfeed_range

  • Sets the duration of active crossfeed during development. High values enforce strong structural similarity.
  • The value x ranges from [0,1], and the crossfeeding is deactivated after x*num_inference_steps steps

branch1_crossfeed_decay

  • Sets decay for branch1_crossfeed_power. Lower values make the decay stronger across the range.
  • The value x ranges from [0,1], and the branch1_crossfeed_power is decreased until the end of the branch1_crossfeed_range to a value of x*branch1_crossfeed_power

parental_crossfeed_power

Similar to branch1_crossfeed_power, however applied for the images withinin the transition.

parental_crossfeed_range

Similar to branch1_crossfeed_range, however applied for the images withinin the transition.

parental_crossfeed_power_decay

Similar to branch1_crossfeed_decay, however applied for the images withinin the transition.

guidance_scale

  • Higher guidance scale encourages the creation of images that are closely aligned with the text.
  • Lower values are recommended for the best results in latent blending.

guidance_scale_mid_damper

  • Decreases the guidance scale in the middle of a transition.
  • A value of 1 maintains a constant guidance scale.
  • A value of 0 decreases the guidance scale to 1 at the midpoint of the transition.

num_inference_steps

  • Determines the quality of the results.
  • Higher values improve the outcome, but also require more computation time.

nmb_trans_images

  • Final number of images computed in the last branch of the tree.
  • Higher values give better results but require more computation time.