Range Scale Float (VixUI)

VixUiRangeScaleFloat looks almost identical to VixUiRangeFloat - same value/min/max/step sliders, same FLOAT output - but the Visionatrix team built it for one specific job: scale factors, as in "how much do we upscale the input image?" Where a strength slider is about strength, this one is about size multiplier, and it's tuned and labeled accordingly (the default display_name is even "Image Size Factor").

In plain ComfyUI it's again a pure pass-through. Its purpose is Visionatrix: a user picks "2.0" and your flow upscales the input by 2x, without them touching a node graph. If you're building an upscale flow or an img2img flow where the input image size matters, this is the control you want instead of the generic Range Float - partly because the label communicates intent, partly because of the extra field it carries.

Inputs that matter

• value (FLOAT, default 4.0) - the current factor. Note the default is 4.0, i.e. 4x, which is a heavy default for an upscaler; you'll probably want 1.0 or 2.0 depending on your model.
• source_input_name (STRING, default "") - this is the field that makes it special. It names the input (e.g. the image input of your flow) whose size this factor scales. Visionatrix uses it to know what the factor applies to. Leave it empty and the node still works as a plain float, but you lose the semantic hook.
• min (default 1.0), max (default 9.0), step (default 0.1) - the slider geometry.
• display_name, optional, advanced, order (default 99), custom_id - the standard VixUI metadata.

hidden is the optional input.

Outputs

One FLOAT output carrying the factor. Feed it into the scale input of your upscale node - UltimateSDUpscale's scale, an ImageScale factor, a tiled upscale multiplier - and your flow scales by the slider.

How it works

Same as every VixUI node: the value passes through untouched and the surrounding metadata is what Visionatrix renders into a control. The only real difference from Range Float is intent and that source_input_name field, which lets the platform tie the factor to a specific input image.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/Visionatrix/ComfyUI-Visionatrix

Restart ComfyUI, or install ComfyUI-Visionatrix via ComfyUI Manager. No model downloads; deps are torch, pillow, numpy.

Gotchas

Two things to watch. First, the 4.0 default: publish an upscale flow without touching it and every image gets 4x'd out of the box, which will melt VRAM on big inputs - set a sane default and a sane max. Second, source_input_name must match an actual input name in your flow or the semantic link silently does nothing. And if you're not targeting Visionatrix at all, skip this node - a plain float input does the same thing with less ceremony.