EmAySee Image Text Preview Passthrough

The caption preview you didn't know you wanted

Every captioning, tagging, or dataset-prep workflow hits the same wall: the image renders in the preview pane, but the text that goes with it lives in a separate node's widget, and you have to squint at two places at once to check whether the caption actually matches the picture. EmAySee_ImageTextPreviewPassthrough fixes that in the laziest possible way - it shows you both in one place and otherwise does absolutely nothing. That's the point.

It's an output node from the EmAySee pack, and its whole job is to render your image in the UI and print the text beside it, while passing both out the other side unmodified. You can hang it at the end of a chain and keep using its outputs for the next node, which makes it handy mid-workflow for QA rather than just as a final "save" sink. If you're building a loop that captions a batch of images, dropping this in lets you eyeball image-to-caption alignment without breaking the flow.

How it actually works

Under the hood it's a passthrough with a side effect. When it runs, it writes the image to ComfyUI's temp folder as a PNG with a random 8-character suffix baked into the filename prefix, then tells the UI to display that file plus the text. The random suffix matters: it means every run generates a fresh file, so your preview doesn't get overwritten by the previous one mid-batch. Then it returns (image, text) - the exact tensors and strings you fed it.

Inputs and outputs

• image (IMAGE) - any image tensor, straight in and straight out.
• text (STRING, multiline) - the caption or label you want displayed alongside it.

Outputs: image and text, identical to the inputs. Wire them onward if you like, or leave them dangling.

Install

Part of the ComfyUI_EmAySee_CustomNodes pack:

cd ComfyUI/custom_nodes
git clone https://github.com/EmAySee/ComfyUI_EmAySee_CustomNodes
# restart ComfyUI

Or use ComfyUI Manager → Install Custom Nodes → search "ComfyUI_EmAySee_CustomNodes". There's no requirements.txt, so nothing extra to pip install - this node runs on PIL/numpy/torch that ComfyUI already ships. Expect a loud magenta "SPECTRE v5.0" ASCII banner in the console on startup; that's normal.

Where people get tripped up

The big one: this writes a temp PNG every run, and the random filename means ComfyUI won't naturally dedupe them. Run a 500-image caption loop and you'll have 500 temp files sitting in ComfyUI/temp/. They're in the temp dir so they're safe to clean, but don't be surprised by the accumulation. Also, since it's an OUTPUT_NODE, it marks the workflow as "done" - if you hang it mid-graph expecting later nodes to still run their previews, the UI treats this as a terminal point. It's a preview-and-check node, not a production sink; for actual saving you want the pack's EmAySee_SaveImage or the core SaveImage.