Krita Strength

Krita Strength is the most niche node in this pack, and it's honest about it - the pack's own metadata calls these "custom nodes for weird niche needs when working with Krita." If you're driving ComfyUI from Krita's AI diffusion plugin, you've met the plugin's strength slider: roughly, 0 keeps your original image, 1 regenerates fully. This node converts that idea into something ComfyUI's samplers actually understand: a step index into the sigma schedule.

What it computes

Samplers work down a list of sigmas - the noise levels a scheduler hands it. If a schedule has N sigmas, there are N - 1 sampling steps to take. This node takes your strength and denoise and figures out which step the run should start at:

• denoise trims the schedule first - round((N-1) * denoise) becomes the number of steps you actually take. Denoise 1.0 means all of them.
• strength then picks a point in that trimmed range - round(effective_steps * (1 - strength)). Strength 0.0 → step 0 (start from the top, full denoise); strength 1.0 → the end (barely touch the image).
• The result is clamped to 0..N-1 so it can't wander off the schedule.

Inputs and outputs

• sigmas (SIGMAS, required) - comes from a scheduler node, the same sigma output a sampler uses. It needs at least two entries; if you feed it a degenerate schedule, the node prints a warning and returns 0.
• strength (FLOAT, default 1, range 0–1) - the Krita-style strength value.
• denoise (FLOAT, default 1, range 0–1) - how much of the schedule to run.

Output: Step (INT) - the index where sampling should start.

Where the Step int goes depends on your workflow - into a sampler that accepts a starting step, or into any node that wants a position in the schedule. If you're using this, you already know where it plugs in; the point is that the math stops being something you approximate by hand.

Honest notes

This is a single-purpose helper, not a general tool. If you don't work in the Krita AI plugin world, you don't need it - there are simpler ways to control denoise in plain ComfyUI. If you do, it saves you from re-deriving the same arithmetic every time.

Two small annoyances: the node prints a debug line to the console on every run (harmless, slightly noisy), and because strength and denoise are both clamped 0–1, you can't use it to push past the schedule for creative over-sampling. That's a feature, not a bug - it keeps the output safe.

Installation

It's in the ComfyUI-Jax-Nodes pack. ComfyUI Manager: search "ComfyUI-Jax-Nodes", install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/Kebolder/ComfyUI-Jax-Nodes

No pip dependencies, no model files, nothing to configure. It uses ComfyUI's newer extension API, so a current ComfyUI build (late 2025 or later) is expected. Small pack, lightly maintained - for a math helper, that's the right amount of attention.