Image Scale Down By
Shrink an image by a factor, no math
- images
- IMAGE
You've got a 2048px render and you need it smaller for the next step. That's the whole job here. Feed it an image, give it a multiplier, get a smaller image back. No width/height fields, no aspect-ratio arithmetic, no upscale model - just "make this 0.5x" and it does.
It's one of the little quality-of-life nodes that make ComfyUI-Easy-Use what it is. The pack is an efficiency bundle built on top of TinyTerraNodes, and half its value is exactly this kind of thing: a job you can do with core ComfyUI, wrapped so you don't have to think about it. Core has an "Upscale Image By" node that scales in either direction; this one is deliberately down-only, which is a nice guardrail when shrinking is genuinely all you mean to do.
Why you'd reach for it
The common cases are all about feeding the next node something at the right size. Generate at high res, then scale down before saving a web-friendly copy. Shrink a reference photo before it goes into a ControlNet preprocessor or an IPAdapter, so you're not burning VRAM encoding a huge image the model will squash anyway. Make a quick thumbnail. Downsample a batch before a grid preview. Anywhere you'd otherwise reach for a resize node and start doing multiplication in your head, this is faster.
How it works
It multiplies both dimensions by your factor and resamples. A 1024x1024 image at scale_by 0.5 comes out 512x512; at 0.25 it's 256x256. Aspect ratio is preserved because both sides get the same multiplier. That's it - there's no separate resampling-method dropdown exposed, so you take the pack's default rather than picking bilinear vs lanczos yourself. For downscaling that's rarely something you'd want to fiddle with anyway.
The inputs that matter
There are only two, and one of them you'll barely touch:
images- the image (or batch) to shrink.scale_by- the multiplier, default 0.5, range 0.01 to 1.0. This is the whole control. Note the ceiling: it maxes out at 1.0, so you literally cannot upscale with this node. That's the point - if you set 1.0 you get the image back unchanged.
The output is a single IMAGE that wires straight into whatever comes next: a Save Image, a preprocessor, an encode node, another scale step.
Installing it
Grab the whole pack through ComfyUI Manager - open Manager, search ComfyUI Easy Use, install, restart. Prefer the command line? cd ComfyUI/custom_nodes && git clone https://github.com/yolain/ComfyUI-Easy-Use, then install the requirements (double-click install.bat on Windows, or pip install -r requirements.txt inside the pack folder) and restart ComfyUI. This particular node has no model downloads and no heavy dependencies - it's pure image resampling.
Where people trip
The one genuine gotcha is the direction: it only shrinks. If you came here trying to enlarge an image, you want core ComfyUI's "Upscale Image By" (which accepts factors above 1.0) or a proper upscale-model node - this one caps at 1.0 on purpose and will silently do nothing at that value. Beyond that, there's nothing exotic to break. If a whole batch of Easy-Use nodes suddenly misbehaves after you update ComfyUI, that's the pack lagging a frontend change, not this node specifically - update Easy-Use and it sorts itself out.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| scale_by | FLOAT | 0.500.01–1 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |