Text Condition Switch (YC)

Compare two strings; if they match, pass image_a through, otherwise pass image_b. It's the image-routing cousin of the pack's Mask Condition Switch - same string-comparison core, different data type, plus a boolean output so you can branch on the result elsewhere in the graph. If you've seen one, you've seen the pattern; the interesting part is where you'd actually put it.

What it's actually for

Per-iteration image selection in batch or loop workflows. The workflow carries text per pass - a filename, a tag, a value from a CSV or wildcard - and the image you want to feed downstream depends on that text. "If the current tag says day, use the daylight version; otherwise use the night version." That's this node. It's also a cheap way to force a branch: match on something and swap the entire input image for a placeholder, a mask preview, or a different crop.

How it works

Exact string comparison. input_text is tested against preset_text, with case_sensitive deciding whether case matters. On match, output image_a; otherwise image_b. is_matched carries the boolean regardless, so you can use the node as a pure text comparator even when you don't care about the image output.

The empty-branch behavior is the thing to remember: if the branch you need isn't connected, it returns a 1×1 black image rather than failing. That keeps the graph alive but can silently wreck downstream nodes that assume real dimensions. And the comparison is exact - trailing whitespace or a line break makes a visibly identical string fail the test.

Inputs and outputs

• input_text - the text under test
• preset_text - what it's compared against
• case_sensitive - default on
• image_a - passed when matched
• image_b - passed when not matched

Outputs: the chosen image and the is_matched boolean.

Install

Part of ComfyUI-YCNodes. ComfyUI Manager → search "ComfyUI-YCNodes" → install, restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/yichengup/ComfyUI-YCNodes

Standard deps (torch, numpy, pillow, opencv-python, scipy), no model downloads.

Where people get burned

• The 1×1 black fallback. Both branches optional, and an unconnected one silently returns a 1×1 image. If your downstream suddenly sees a black pixel instead of a photo, this is the usual suspect.
• Exact match, spaces and all. The comparison is ==, not "contains" and not trimmed. Sanitize text before it arrives, or matches you expect will quietly fail.
• Expecting it to branch more than two ways. Two inputs, one test. For many-way image selection, use an index switch instead.

It's a two-image router with a string condition - a small piece, but the kind of thing that turns a manual "swap the input by hand" workflow into one that decides for itself.