EZ txt2img

The node's class is EZBlank, but ComfyUI's menu shows it as EZ txt2img - which is the name you'll actually look for. And it is precisely the txt2img entry point of the Prompt Helpers pack: "generate a random image from nothing" in the pack's own words. Two inputs (width, height), one settings output, and the entire rest of the pipeline hangs off it.

The confusion about the name is worth flagging up front because it's the single most common search miss. You're not looking for "blank canvas" - you're looking for "text to image," and this is the node that says "start from empty latents rather than an existing image."

Inputs

• width (INT, default 1024, step 8, range 16–16384) - image width in pixels.
• height (INT, default 1024, step 8, range 16–16384) - image height in pixels.

The step of 8 is deliberate: latent-space images need dimensions that are multiples of 8 (ComfyUI enforces this everywhere, and the pack snaps crop/detail sizes to it too). Leave the defaults alone unless you have a reason - 1024×1024 is the SDXL-family home base, and going to weird resolutions is a common source of "why does this look off" that isn't actually the node's fault.

The output is IMAGE (EZ_IMAGE_SETTINGS) - a settings bundle that says "blank canvas, this size," which then flows into EZ Generate (or through EZ Crop, EZ Detail, or EZ Batch first, since they all operate on the same settings type).

How it works

When EZ Blank executes, it builds a settings dict stamped type: BLANK with your width and height plus defaults for everything else (no crop, no detail, batch 1). It does not create a latent yet - nothing is allocated until EZ Generate runs, at which point it sees type: BLANK and constructs an EmptyLatentImage at exactly your dimensions. That's why the chain feels instant to edit: changing width/height is just editing two numbers, not re-encoding anything.

The shape of a complete workflow

EZ txt2img ──► EZ Batch ──┐
                          ├─► EZ Generate ──► EZ Generate Save
EZ Prompt ────────────────┘
EZ Sampler ───────────────┘

Swap the start for EZ img2img when you have a source image, or add EZ Detail for a face-fix pass - the image settings bundle is the same type throughout, so these compose freely.

Installing it

It ships with the Prompt Helpers pack. In ComfyUI Manager search "Prompt Helpers" (Pauan/comfyui-prompt-helpers), install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/Pauan/comfyui-prompt-helpers

Restart ComfyUI. No model downloads - this node is pure settings, and the pack's only real dependencies are PyYAML and desktop-notifier (used by EZ Notify). Requires ComfyUI 0.8.0+ since the whole pack runs on the newer Nodes 2.0 API.

If you set a resolution and EZ Generate silently refuses or the image comes out a different size, check the multiple-of-8 rule first: a 1025×1024 request snaps weirdly in latent land. Stay on 8s and you'll never think about this node again - which is the compliment it's aiming for.