DP Draggable Int 8step

A draggable integer that only lands on multiples of 8, from 0 to 1000. This is the sibling you want when the value has to be divisible by 8 - which, in diffusion land, is a surprising amount of the time.

Image dimensions are the classic case: latents are downsampled 8x from pixel space, so widths and heights need to be divisible by 8 or the model errors out. A lot of frame counts and batch groupings step in eights too. Using an 8-step knob means you can't accidentally set a 100 where the pipeline needed a 96 or 104 - it snaps to the grid for you, and that's one whole class of "why did my run fail" removed.

The one input

• value (0–1000, step 8, default 500) - drag or type; it resolves to the nearest multiple of 8.

Output is a single value (INT) for any integer socket downstream.

Why 8 specifically

Eight is the mechanical floor for diffusion dimensions - the VAE downsamples by 8, so anything you feed as a width or height has to divide evenly by it. (Many workflows prefer multiples of 64 for cleaner results, but 8 is the hard requirement.) If you're driving a dimension, a batch grouping, or a frame count that comes in eights, this knob keeps you honest without you having to do the arithmetic. It's the same draggable-knob ergonomics as the rest of the family; the value-add is the guardrail.

How to install it

ComfyUI Manager: search ComfyUI-Desert-Pixel-Nodes, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/DesertPixelAi/ComfyUI-Desert-Pixel-Nodes

then restart. No models, no dependencies. Under the "DP" utils category.

Common issues & troubleshooting

I need an odd or non-8 value. This node won't give you one - snapping to 8 is the whole feature. Use the 1-step draggable int for every integer, or the 4-step for multiples of 4.

I need bigger than 1000. Capped at 1000. For a larger dimension or a longer frame sequence, the DP Diff Int 8step Selector runs to 1920 on the same multiple-of-8 grid, or use a core int node. That said, pushing base resolution far past your checkpoint's native size tends to duplicate subjects - upscale in a later pass instead of cranking the dimension here.

Which stepped int do I want? Match the step to the requirement: 8 for dimensions and most frame counts, 4 for values that group in fours, 1 when any integer is fine. Picking the right grid up front means the number literally cannot land somewhere the downstream node will reject.