ComfyUI-Advanced-Latent-Control

This custom nodes helps to transform latent in different ways.

You can access new features earlier by switching from the master branch to dev,

but you need to remember that there may be some issues on the dev branch and some nodes’ behavior may change after release.

Latent mirror

This node can flip latent and merge original and flipped version.

Input:

latent

Fields:

direction – can be vertically, horizontally or both

multiplier – multiply latent by specified number

Output:

latent

Usage:

Latent shift

This node can shift latent along x and y-axis.

Input:

latent

Fields:

x_shift – a number between -1 and 1 that indicates how much the latent should be shifted

y_shift – a number between -1 and 1 that indicates how much the latent should be shifted

Output:

latent

Usage:

~~TSampler with transforms (Latent Control)~~

Removed from version 2.0.0

TSampler (Latent Control)

This node allows to combine a lot of transforms with different parameters.

Input:

base KSampler fields

transform_optional – field that can take output from one of those nodes: Mirror transform, Shift transform, Multiply transform or Combine transforms

Fields:

exactly matches the base KSampler

Output:

exactly matches the base KSampler

Usage:

Multiply, Mirror and Shift transform nodes parameters exactly match the corresponding KSampler with transforms (Latent Control) parameters.

There are two new transform nodes:

Latent add

Latent interpolate

They work exactly the same as LatentAdd and LatentBlend nodes from standard node pack, but also, can multiply result by specified number.

Offset

You can apply specific offset for transform nodes.

Fields:

process_every – a number that indicates which steps will be processed

offset – a number that indicates offset for previous parameter. For example: if process_every is 4 and offset is 0, sampler apply transformation with this pattern: 0 0 0 1. This pattern will repeat again and again. If offset is 2, pattern will be 0 1 0 0, if -1 – 1 0 0 0.

mode – can be process_every or skip_every. For example, with skip_every previous pattern (0 0 0 1) turn into this: 1 1 1 0

Output:

offset

Usage:

You can combine different offsets to achieve interesting patterns. For example:

0 0 0 1 and 0 0 1 give this pattern: 0 0 1 1 0 1 0 1 1 0 0 1.

One time nodes

Each transform node has own one-time version. They allow to make one transform action at specified step.

Usage:

Latent normalize

Fixes some issues when sampling modified latent space.

Input:

exactly matches the VAE Decode node

Output:

latent

When you multiply latent by negative or big positive (bigger than 2) number and paste this latent in sampler, you can see that the

image will be generated very poorly. This is because stable diffusion cannot work with such set of numbers (meaning the numbers contained in latent).

But you can prevent this behavior by sequential decode and encode latent using vae. Node Latent normalize make this process easier.

This node also change some results even if output without this node looks good.

And it very slightly changes results from latent, which have not been modified.

Transform hijack

Allow you to use transforms with any samplers that you like.

Inputs:

latent

transforms

Outputs:

latent

Usage: