LatentUpscaleFactorSimple _O

The two-factor version of this pack's latent upscaler (LatentUpscaleFactor _O) is great until you realize you set the same number in both boxes 95% of the time. LatentUpscaleFactorSimple _O is the version that assumes you want a uniform scale: one factor input, applied to both width and height. 1.25 in, 1.25× both dimensions out. It's the same machinery with one less widget.

It's part of the O/latent folder of the Quality of Life Suit pack, and honestly it's the one I'd reach for first. When you're doing a straightforward "bump this latent up a bit" - the classic hi-res-fix intermediate step - a single multiplier is all you want, and two boxes is two places to make a typo.

How it works

Identical mechanics to the two-factor node: it reads the latent's dimensions, multiplies both by factor, and runs ComfyUI's common_upscale with your chosen method. Same good behaviour inherited along the way:

• Multiples of 8 enforced - output dimensions are rounded to the nearest multiple of 8, which SD-family models need.
• 8192 cap - a runaway factor can't produce an absurd tensor.
• Before/after sizes printed to the console - upscale from (512,512) to (640,640) so you can confirm what you got.

Inputs:

• samples (LATENT) - the latent to scale.
• upscale_method - nearest-exact, bilinear, or area.
• factor (FLOAT) - the single multiplier, default 1.25, range 0–10. Values below 1 downscale.
• crop - disabled or center.

One output: LATENT.

Where it fits

Same honest advice as its sibling: a latent upscale is a cheap intermediate step, not a true upscaler. Keep the factor modest (1.25–2×) when you plan to follow with a second sampling pass, and don't expect latent interpolation to add crisp detail - that's what pixel-space upscalers and a low-denoise second pass are for. Where this node wins is iteration: drag one knob, queue, compare. If you never need per-axis factors, this is the one to use and the two-factor version is dead weight.

Installing it

Part of the Quality of Life Suit pack (omar92/ComfyUI-QualityOfLifeSuit_Omar92). Install via ComfyUI Manager (search "Quality of Life Suit") or clone:

cd <ComfyUI>/custom_nodes
git clone https://github.com/omar92/ComfyUI-QualityOfLifeSuit_Omar92.git

Restart ComfyUI properly. No models, no dependencies.

Common issues

• "The size isn't exactly factor × input." Multiple-of-8 rounding again. Expected.
• "Output looks soft." Latent interpolation is inherently soft - pair it with a second-pass sampling at low denoise for the classic hi-res fix.
• "I actually need different X and Y factors." Then you want LatentUpscaleFactor _O instead. That's the only real difference between the two nodes.

One knob, both dimensions, zero thinking. That's the entire pitch, and for a lot of workflows it's all you need.