ConditionTextPrompts

ConditionTextPrompts is the stripped-down sibling of this pack's Prompts node - same core idea, positive and negative CLIP encoding out of one node instead of two, but leaner. Where Prompts gives you multiline fields with placeholder defaults, echoes the text back out as strings, and treats clip as optional, ConditionTextPrompts uses single-line fields, requires clip up front, and only gives you the two CONDITIONING outputs - no text passthrough.

If you're deciding between the two: reach for Prompts first, especially as a beginner - the multiline fields and echoed text outputs make it more forgiving and more useful for things like stamping your prompt onto an image afterward. Come to ConditionTextPrompts when you want the minimal version and don't need anything Prompts adds - one less thing on the node, nothing else changes.

How it works

It's two CLIPTextEncode calls bundled together, nothing more exotic than that. You type a positive and a negative, feed in a CLIP, and get back two independent CONDITIONING tensors - the same math ComfyUI's own encoder does, just packaged so you don't place and wire two separate nodes for what's conceptually one step (write your prompt pair).

The inputs and outputs that matter

• clip - required. Connect it from your checkpoint loader's CLIP output.
• positive - single-line text, defaults empty.
• negative - single-line text, defaults empty.

Three outputs: CLIP passes the model through so you can chain another text node without a second wire back to the loader; positive and negative are CONDITIONING, ready for your KSampler's two conditioning sockets.

How to install it

• ComfyUI Manager: search "ComfyUI-Chibi-Nodes", install, restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/chibiace/ComfyUI-Chibi-Nodes/, restart.

No models, no extra dependencies - pure logic over ComfyUI's own CLIP encoding. It lands under Chibi-Nodes/Text in the menu. The pack as a whole is a small personal utility collection - the README itself says "experimental nodes" and admits version 2 shipped "with more bugs" - which for a node this simple mostly just means: don't expect polish beyond what does the job.

Common issues

Because both text fields are single-line, this node is a poor fit for long, detailed prompts - you'll be scrolling a cramped widget rather than reading your prompt comfortably. If your prompts run more than a short phrase, either use the Prompts node instead (multiline by default) or write your text in a Textbox node and feed it into a plain ConditionText node's input, since ConditionTextPrompts doesn't accept STRING inputs for positive/negative - they're fixed widgets, not sockets, unless you manually convert them.

The other thing worth flagging since it's easy to miss with two nearly-identical nodes in the same pack: don't mix up which one you dragged in. If you've got a graph with placeholder text still reading "Positive Prompt" / "Negative Prompt" sitting in your generation, you've grabbed Prompts, not this one - ConditionTextPrompts defaults to empty fields, so an unedited run here produces unconditioned output on both sides, not a visible placeholder to warn you.