marduk191’s Image manipulation with unsampler (Textual Inpaint)

This is a workflow tool to manipulate existing images using the unsampler node. Basic instructions with basic descriptions are provided in the workflow. This is compatible with any model that has a loader and ksampler available in ComfyUI.

Tip jar

What this workflow does

Provides a pre existing image manipulation technique using the unsampler node.

How to use this workflow

Positive Manipulation – A summary of your image. The more you deviate this from the original image, the more Render weight you will need.

Negative Manipulation – Not usually needed, works as usual though, this will have undesirable results in many cases.

Checkpoint – Load a model with the style suitable for your manipulation.

Starting Image – Image you wish to manipulate

Aspect Ratio – Picking an aspect ration closely matching your starting image is recommended.

Tips about this workflow

Sampler Name – Sampler used for decoding and encoding latents. This must match so that the latents are compatible.

Step Switch – This specifies where to stop converting the base image, and where to start rendering the latents created at that step.

Scheduler – Basic noise sampler, This must match so that the latents are compatible.

Total Steps – Total sampling steps of the unsampler, and sampler. I made this global, you can change it, It is not recommended.

Unsample CFG – Weight of the base image when converting to latents. (Likeness of original)

Render CFG – Weight of the checkpoint model when rendering modified latents. (How much Help we need from the model)