Preview Image + Text 📝

Preview Image + Text is the pack's answer to a dumb, persistent problem: in a randomize-and-freeze workflow, you're generating variations and you need to know which prompt produced which image. The stock Preview Image shows you the picture; this node shows you the picture with the text that was behind it sitting right there in the node. No squinting at the workflow, no mental bookkeeping across ten random selections.

How it works

Under the hood it subclasses ComfyUI's own SaveImage, but pointed at the temp preview directory - same as the stock Preview Image node, so it writes to the preview folder rather than your permanent output folder. On every execution it saves the incoming images, then pushes the current text back into the text_display widget so the label updates in sync with the picture. That "sync" is the key word: the text updates only when the node executes, which means the label always reflects the run that produced the image you're looking at.

Two text paths exist. text (optional) is a wire input - connect a Random Selector, a combiner, or any string output here and that value is displayed. text_display is the widget fallback when nothing is wired in. If both are set, the wire wins. The node has no outputs; it's a terminal preview, which is exactly what you want at the end of a generation pipeline.

The inputs that matter

• images - the IMAGE tensor you want to preview. Wire this from your VAE decode.
• text (optional) - the live value to show. Wire your prompt string here.
• text_display - the widget showing the current text; edit it only if you want a static label.

Gotchas

It writes to the temp directory, not ComfyUI/output. The comfy.icu listing describes it as saving to your output directory, but the shipped code sets the temp preview folder - so don't go hunting for these images in your regular output folder. They're the same transient previews the stock node makes. And since it's a preview node, unplugging it from the graph entirely is harmless to your actual saved outputs.

How to install

The whole pack installs with no dependency wrangling - pure Python stdlib plus ComfyUI's own classes, zero model downloads:

cd ComfyUI/custom_nodes
git clone https://github.com/Ckrest/comfyui-easy-randomize.git

Restart ComfyUI, or install "comfyui-easy-randomize" from ComfyUI Manager. It registers under Local Custom/Output (the README's "Easy Randomize" category doesn't exist in the shipped code).

Where it fits

This is the review step of the pack's core loop: Randomize → Preview → Freeze. Run a few batches, glance at each preview with its selection readout, check the frozen boxes on the combos you like, re-run. If you want the text baked into the file rather than shown on a preview, use the pack's Save Image (with Selections) instead - this node is for the human eyeball phase, and it does that job without any ceremony.