ConditionText

ConditionText is the simplest of the three "condition text" nodes in this pack - plain text in, CONDITIONING out, same as ComfyUI's built-in CLIPTextEncode. The one difference that matters: it also passes the CLIP model straight back out as a second output, so you can chain it into another text-encoding node downstream without running a separate wire back to your checkpoint loader every time.

That sounds minor until you're building a graph with several text-encode steps - say, a positive prompt, a negative prompt, and a couple of regional or LoRA-trigger conditionings all needing the same CLIP. Daisy-chaining CLIP out → next node's clip in keeps the graph readable instead of four wires converging on one checkpoint loader from across the canvas.

The inputs and outputs that matter

• clip - required, feed it from your checkpoint loader (or from another ConditionText/Prompts node's CLIP output, if you're chaining).
• text - required. Note there's no multiline flag set on this field in the node's own schema, unlike Textbox or Prompts, so expect it to render as a plain single-line widget. For anything longer than a short phrase you'll want to type it in a Textbox node instead and feed it in as a STRING, rather than fighting a cramped single-line box.

Two outputs: CLIP (the passthrough) and CONDITIONING, which wires straight into a KSampler's positive or negative socket - this node doesn't know or care which; that's determined entirely by where you plug it in.

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 dependencies - this is pure node logic layered over ComfyUI's own CLIP text encoding, so it inherits whatever encoder your checkpoint uses. It shows up under Chibi-Nodes/Text once installed.

Worth knowing for the pack as a whole: the author's README describes these as "experimental nodes," and jokes that version 2 arrived "with more bugs." For a node this thin - it's barely more than a wrapper around ComfyUI's own encode call - that's not much of a risk, but it does mean don't expect maintained docs beyond what you're reading here.

Common issues

If you're coming from a workflow that used the stock CLIPTextEncode and switched to ConditionText expecting identical behavior, you should get it - the conditioning math is the same, it's the CLIP passthrough that's new. If your output doesn't look like your prompt is landing the way you expect, the usual suspects apply and have nothing to do with this node specifically: check you're feeding the right CLIP (the one matching your checkpoint, not a mismatched one from a different model loaded elsewhere in the graph), and remember that CLIP-skip settings (like Chibi's own Loader node's stop_at_clip_layer, or ComfyUI's native CLIPSetLastLayer) need to sit upstream of this node to take effect - ConditionText just encodes whatever CLIP arrives, it doesn't modify it.

One easy trip-up specific to this node: because text has no multiline flag, pasting a long prompt in can end up scrolled off to one side in a narrow single-line box, making it easy to miss that you've accidentally truncated or duplicated part of it. If your prompts run long, prefer Textbox feeding into this node's text input over typing directly into the widget.