VELVET VICE KREA — Create Base Preview

If you've loaded the Velvet Vice Krea 2 Vision Prompter workflow and are poking around the CREATE branch, this is the little node that lets you actually see the base image before it gets refined. Its real name is "VELVET VICE KREA - Create Base Preview," and despite the scary Passthrough suffix it is exactly as simple as it looks: an image goes in, a preview is saved, and the same image comes right back out. No rerouting, no mode logic, nothing to configure.

Why it's there

Krea 2's CREATE studio in this workflow is a pipeline: base (first sampler) → mild refinement (second sampler) → optional detailer → SeedVR2 final pass. When you queue the whole thing, each stage runs in sequence and you only see the end result. This node is the checkpoint in the middle - it shows you what the first sampler produced before refinement starts reshaping it. That's genuinely useful, because it's the moment where you can tell whether your prompt and seed actually landed, before the workflow spends another pass "improving" it.

ComfyUI's native PreviewImage node can do this too, of course. The difference is that this version saves with a fixed VELVET_VICE_KREA_CREATE_BASE_PREVIEW prefix, stamps a status line on the run, and passes the tensor through untouched - so it can sit inline in the middle of the branch without you having to rewire anything around it. The author uses the same trick in a few places in this pack, which is why the whole family of ...Preview nodes exists.

Inputs and outputs

It has exactly one input and one output, both IMAGE:

• images - the rendered base image coming out of your first KSampler. This is the only thing you set up.
• image - the same image, unchanged, ready to feed into the refinement sampler or anything else downstream.

That's the whole interface. If you're on a different branch (Native Edit or Classic Img2Img), this node isn't for you - it's a CREATE-specific passthrough and it doesn't take a workflow mode input to check against. The router-family nodes in the same pack handle cross-branch selection instead.

Installing and running

It ships inside the velvet-vice-krea pack, so you get it with the whole bundle:

cd ComfyUI/custom_nodes
git clone https://github.com/Velvet-Vice/velvet-vice-krea

Then restart ComfyUI. Or search "VELVET VICE - KREA" in ComfyUI Manager. No pip dependencies and no bundled model files - the pack needs nothing extra to install, though the Krea 2 diffusion model it previews still has to be in your diffusion_models folder (more on that in the Model Loader node).

Troubleshooting

The only real failure mode here is a missing ComfyUI PreviewImage core node, which would mean your ComfyUI install is broken in a way that affects every preview node, not just this one. Beyond that: if the preview shows up but you wanted the file, it's in your ComfyUI output folder under the VELVET_VICE_KREA_CREATE_BASE_PREVIEW prefix. And if the node reports nothing at all, check that the sampler feeding it actually ran - this is a passthrough, so no image in means no image out. It can't invent a render for you.