LTFloat_Step_05

LTFloat_Step_05 is the 0.5 member of the Latent Tools step-helper family, and it exists for the sweeps where you want every image to be obviously different from the last. It's a calibrated constant: type a value, get the same number back as a FLOAT and a STRING, with the widget stepping by 0.5.

The family's purpose is parameter sweeps with a locked seed - change one variable per run, compare. At 0.5 per run you're no longer mapping a curve so much as sampling it. A mean sweep from -1 to 1 is five runs that will look wildly different from each other. That's the point: this is the step for demonstrating that a parameter matters, or for grabbing a few representative points across a big range. It is not the step for precision or smoothness - for that you drop to 0.05 or 0.01.

How it works

Straight talk about the mechanism: the node does not auto-increment. It takes the value you give it, rounds to five decimals, and returns it as a FLOAT and a STRING. The step is the widget's spinner increment and the label for the constant. You bump the value between runs; the node keeps the process deliberate.

The inputs and outputs

• value - FLOAT, default 0, range ±1,000,000, step 0.5.
• FLOAT - the value.
• STRING - the value as text, for filename prefixes so each output records its parameter.

FLOAT into a mean/std/cfg input, STRING into a Save Image filename, seed fixed, and you'll see exactly how much each half-unit changes the image.

Installing it

Part of xl0's Latent Tools pack. ComfyUI Manager → search Latent Tools → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/xl0/latent-tools

Restart, and it's under LatentTools. The pack's only dependency is lovely-tensors (Manager installs it; manual cloners may need pip install lovely-tensors). No model downloads.