Image Empty

Sometimes you don't need a generated image - you need a blank one. Image Empty (ImageEmpty) from the zsq_prompt pack creates a solid-color image tensor at whatever size you ask, in whatever color you ask. That sounds trivial, but it's genuinely useful when you want a canvas to composite onto, a placeholder to keep a graph flowing, or a colored background to paste content into.

How it works

You give it a width, height, and batch_size, plus RGB values in color_r, color_g, color_b. It builds a tensor filled with that color and returns it as a standard IMAGE. The color is normalized to the 0–1 range ComfyUI expects internally, but you think in 0–255 when setting the sliders. Defaults are 512×512 and pure black.

That's the whole mechanism. No diffusion, no VAE, no model in memory - it's instant and costs nothing.

Inputs and outputs that matter

• width / height - canvas size, 1 to 16384. The obvious one.
• batch_size - how many identical color frames to produce (up to 4096). Keep it at 1 unless you specifically need N copies.
• color_r / color_g / color_b - the fill color, each 0–255. Black by default; bump them for a colored backdrop.
• Output: image - a solid-color IMAGE tensor ready to wire anywhere an IMAGE goes.

Where you'd use it

• Composite base. Paste a cropped face or a generated element onto a plain background instead of compositing onto nothing.
• Placeholder / canvas. Keep a workflow alive while you decide what actually generates content; blank out a preview while you iterate.
• A cheap mask-ish layer. It's a solid RGB image, not a mask, but for "everything here is one value" needs it does the job before you convert it.
• Batch filler. Need five black frames as a starting point for something? Set batch_size to 5 and move on.

It pairs naturally with the pack's other image tools - generate the blank, then use imageCrop or imageRotate on it, or feed it as the base for a concat/composite step.

Installing zsq_prompt

This is one of ~35 nodes that arrive together with the pack. In ComfyUI Manager, search zsq_prompt and install, or from a terminal:

cd ComfyUI/custom_nodes
git clone https://github.com/windfancy/zsq_prompt

Restart ComfyUI after. The pack installs a genuinely heavy dependency set (transformers, opencv-python, opencv-contrib-python, scikit-image, timm, onnx, ultralytics) even though this particular node only needs numpy/torch. That's just how grab-bag packs work - pay the install once, use the utilities. If Manager ever shows "conflicting with zsq_prompt" on a workflow you load, that's dependency-overlap noise; it's only meaningful once the pack is actually installed. No model files needed for this node.

Gotchas

• It's a solid fill, not transparent. If you're after an alpha channel you'll need to convert elsewhere.
• The output is RGB, so anything expecting a MASK will reject it until you convert.
• Step is 1 pixel on width/height - no forced multiples of 8, which is occasionally useful when you want an exact odd size for compositing.