Image & Text Switch🧸

Same idea as the MultiLine Text Switch, but it carries an image along with the text. You give it up to three text-and-image pairs, set a dial to 1, 2, or 3, and both the text and the image from that slot pass through together. It's the two-lane version, for when your prompt variants each come with their own reference image and you want them to stay matched when you switch.

Small utility node, no AI or key involved. It just routes.

How it works

Each slot is a pair: Text_1 with Image_1, Text_2 with Image_2, Text_3 with Image_3. You set active_input to the slot you want, and the node sends that slot's text out the Multiline Text output and that slot's image out the Image output, in lockstep. So switching to slot 2 gives you text 2 and image 2 together, no chance of them drifting out of sync.

The inputs

• active_input (1 to 3, default 1): picks which pair is live.
• Text_1 / Text_2 / Text_3 (multiline): the text for each slot.
• Image_1 / Image_2 / Image_3 (IMAGE): the image for each slot.

Outputs: Multiline Text and Image, the pair from whichever slot you selected.

Where it's actually useful

The keeping-them-matched part is the whole value. Picture a workflow that runs a vision model (like Style Prompt or Advanced Prompt Enhancer with an image input) where each reference image has a specific instruction that goes with it. Wire each image and its instruction into a matching slot, and one dial swaps the whole setup. Without this you'd juggle two separate switches and eventually flip one but forget the other, which is exactly the kind of silent mismatch that wastes a render. If you only need to switch text, use the plain MultiLine Text Switch instead; there's no reason to carry an image lane you won't use.

Installing it

Ships in Plush-for-ComfyUI. Install through ComfyUI Manager (search "Plush", install Plush-for-ComfyUI, restart), or clone it: cd ComfyUI/custom_nodes && git clone https://github.com/glibsonoran/Plush-for-ComfyUI.git, then pip install -r requirements.txt, restart.

Common issues

The predictable one: pointing active_input at a slot where you only filled in the text but not the image (or vice versa), so half the output is empty. Make sure the slot you select has both lanes wired, or that downstream nodes tolerate an empty one. Otherwise this is a set-and-forget node with basically nothing to troubleshoot.