PixaromaHelp
Help › Nodes › Values

Sizes Pixaroma

Your own list of favourite resolutions, with a one-click orientation flip.

In ComfyUI: double-click an empty spot on the canvas and type Sizes Pixaroma.

What it does

Keep the exact width x height sizes you use in one list, pick one, and it outputs width and height, plus an empty latent at that size. Wire width and height anywhere a size is needed.

The latent output

The latent is the same as ComfyUI's Empty Latent Image, so you can wire it straight into the KSampler and skip that node. It works for image models: SD 1.5, SDXL, SD3, Flux, Flux 2, Qwen Image and others, because the KSampler adapts it to the model.

That adapting step is in ComfyUI's own samplers (KSampler, SamplerCustomAdvanced). Some samplers from other packs (Easy-Use, Impact Pack) skip it, so with Flux 2, Qwen Image 2.1 and similar models the picture comes out twice the size. With those samplers, wire width and height into the model's own latent node.

The batch box on the same line as the latent output sets how many pictures one Run makes, from 1 to 64: click + or -, or click the number and type. It only changes the latent; width and height stay the same.

For video models, wire width and height into that model's own latent node instead. For Flux 2, keep the size a multiple of 16 (turn on Snap 16 in settings) to get exactly the same size as its own latent node.

Portrait and Landscape

The two buttons flip the chosen size between tall and wide. Add a size like 1024 x 1536 once and reuse it in either orientation - you do not need a separate entry for each. Square sizes look the same both ways.

Settings

Click the gear (or right-click the node) to open the settings. There you can add a new size, remove or reorder sizes by dragging, load a set of common sizes in one click, pick the highlight colour, and snap width and height to a multiple of 8, 16, 32, or 64 so the numbers stay friendly for the VAE. Snapping is off by default.

Marking the recommended sizes

Click the star beside a size in settings and it gets a star on the node too. Handy when a list holds every size a workflow accepts but only some of them suit the model you are running: star those, and you can see at a glance which to pick.

A star is a reminder only - it changes nothing about what the node sends out, and you can still choose any size in the list. Stars follow the size itself, so reordering the list, flipping between portrait and landscape, or removing a size and adding it back all keep them.

Labels: find a size faster

In settings, click Tag beside a size and pick 1K, HD, Full HD, 2K, 4K or 8K. The label shows at the left of that size on the node, so you can spot it in a long list. Pick No tag to take it off.

A label is a reminder only, like a star. You choose which size gets which label, it changes nothing about what the node sends out, and it follows the size when you reorder the list or flip the orientation.

The buttons on the node

The fold arrow
Collapses the list down to just the chosen size, or opens it again.
batch, beside the latent output
How many pictures the latent makes in one Run (1 to 64): + and -, or type a number.
Portrait and Landscape
Shows the chosen size tall or wide, swapping which number is the width.
The gear
Opens the settings panel, where you add, remove, reorder and recolour your sizes.
A size in the list
Click any one to make it the size this node sends out.
Add, in settings
Type a width and height, then add it to your list.
The drag handles, in settings
Drag a size up or down to reorder the list.
The star, in settings
Marks a size as recommended. It then shows a star on the node. Click again to unmark.
Tag, in settings
Puts a label like HD or 4K on a size, shown on the node. Pick No tag to take it off.
Snap, in settings
Rounds every size to a multiple of 8, 16, 32 or 64.
Add common sizes, in settings
Adds a set of popular square sizes from 512 up to 2048 in one click.

What comes out

widthINT
The chosen width in pixels.
heightINT
The chosen height in pixels.
latentLATENT
An empty latent at this size: the same as ComfyUI's Empty Latent Image. The batch box beside this output sets how many pictures it makes per run (1 to 64). A KSampler adapts it to most image models.
A fresh node starts with one size (1024 x 1024). Add up to 40 per node.

This is the help that comes with Pixaroma 1.4.251. Inside ComfyUI, select the node and click the orange ? above it to read the same help there.