CLIP Positive-Negative w/Text (WLSH)

Two CLIPTextEncode nodes, one box on your canvas instead of two. That's genuinely the whole pitch - the README calls it "mostly a space saver," which is refreshingly honest for a custom node. If you're tired of your graph having a positive box and a negative box sitting next to each other doing nothing interesting, this collapses them into one.

What makes this specific variant worth picking over the plain CLIP Positive-Negative (WLSH) node is the extra pair of outputs: alongside the two CONDITIONING tensors your sampler actually needs, it also hands back the raw positive_text and negative_text strings you typed in. That sounds trivial until you're wiring up one of WLSH's own save nodes (Image Save with Prompt, Image Save with Prompt File) - those want the actual prompt text as a string input so they can bake it into the saved metadata, and without a text-passthrough node like this one you'd be typing your prompt twice: once here for encoding, once again in the save node so it ends up in the file. This node means you type it once and both destinations get fed from the same source.

Inputs: positive and negative are both multiline STRING boxes - your actual prompt text. clip takes the CLIP model, same as it comes out of your checkpoint loader. Outputs, four of them: positive and negative are CONDITIONING, wire those into your sampler exactly like you would from a stock CLIPTextEncode. positive_text and negative_text are the same strings you typed, passed straight through unmodified - wire those into anything downstream that wants the prompt as text rather than as conditioning.

Worth knowing regardless of which save node you pair this with: your negative prompt only does anything once CFG is above 1. At CFG 1 - which is the default on distilled checkpoints like Z-Image Turbo, Flux Dev/Schnell, and various Lightning/Turbo variants - the field is silently ignored, and ComfyUI doesn't even bother computing the negative pass under most samplers at that setting. If you're on one of those and your negative prompt seems to be doing nothing, that's not this node malfunctioning, that's just what CFG 1 means.

Installing it: search "wlsh_nodes" in ComfyUI Manager, or clone manually:

cd ComfyUI/custom_nodes
git clone https://github.com/wallish77/wlsh_nodes

Restart ComfyUI. No models to download, no extra dependencies - it's a thin wrapper around CLIP text encoding, so nothing beyond the pack itself is required.

Troubleshooting: if your positive and negative results look swapped, double-check the two text boxes weren't accidentally filled backwards - an easy mistake with two nearly-identical multiline fields stacked on top of each other. If positive_text/negative_text aren't accepted somewhere you expected, remember they're plain STRING type - they won't plug into a CONDITIONING socket, and vice versa the positive/negative conditioning outputs won't plug into a text field. It's easy to grab the wrong one of the four output dots when they're all clustered together on a small node.