Nodes/ComfyUI-PolyhedronLoRAStack/⬡ Polyhedron CLIP Text Encode
ComfyUI Node

⬡ Polyhedron CLIP Text Encode

Segmented prompt encoding: the positive prompt is assembled from several fields instead of one wall of text, so a trigger block, a subject and a style block can be edited without hunting through a paragraph. Comments and stray newlines are stripped before encoding, and the text outputs hand back exactly what was encoded - what the encoder saw is what you can read.

By PolyhedronAI·Created 3 months ago·Updated 17 days ago· 2
⬡ Polyhedron CLIP Text Encode
  • clip
  • positive
  • negative
  • positive_text
  • negative_text
  • full_text
segments3
pos_1
pos_2
pos_3
pos_4
pos_5
pos_6
use_negativetrue
neg_1
separatorcomma
strip_commentstrue
strip_newlinestrue
external_modeappend
comment_markers//
pos_external
neg_external
CategoryPolyhedron/Conditioning

Inputs (17)

NameTypeDefaultDescription
clipCLIPThe CLIP / text encoder to use (same pin as the core node).
segmentsINT31–6How many positive segments are visible. Turning this DOWN only hides fields - their text stays saved.
pos_1STRINGPositive segment 1. Keep your sections apart for readability - they are joined into ONE prompt by the separator. Hidden segments keep their text.
pos_2STRINGPositive segment 2. Keep your sections apart for readability - they are joined into ONE prompt by the separator. Hidden segments keep their text.
pos_3STRINGPositive segment 3. Keep your sections apart for readability - they are joined into ONE prompt by the separator. Hidden segments keep their text.
pos_4STRINGPositive segment 4. Keep your sections apart for readability - they are joined into ONE prompt by the separator. Hidden segments keep their text.
pos_5STRINGPositive segment 5. Keep your sections apart for readability - they are joined into ONE prompt by the separator. Hidden segments keep their text.
pos_6STRINGPositive segment 6. Keep your sections apart for readability - they are joined into ONE prompt by the separator. Hidden segments keep their text.
use_negativeBOOLEANtrueOff hides the negative box and encodes an EMPTY negative (a valid conditioning - the graph stays intact).
neg_1STRINGThe negative prompt.
separatorCOMBOcommaHow the segments are joined into one prompt (comma = ', ').
strip_commentsBOOLEANtrueRemove // comments BEFORE flattening newlines. Your section headers stay in the box but never reach the model (left in, they would cost tokens). Off = the historic behaviour.
strip_newlinesBOOLEANtrueFlatten line breaks into spaces after the comment pass.
external_modeCOMBOappendWhat to do with a wired external text (e.g. Florence2): add it behind, in front, or let it REPLACE the segments.
comment_markersSTRING//Which markers start a comment (used by strip_comments). SPACE-SEPARATED list, e.g. '// # ***'. A marker counts at the line start or after a space, and everything up to the end of the line goes - so '//Comment' and '// Comment' both work, while 'https://...' survives. Empty = nothing is stripped.
pos_externaloptSTRINGExternal positive text (Florence2, a Stack's trigger words, ...). It shows up in the node preview.
neg_externaloptSTRINGExternal negative text.

Outputs (5)

NameTypeDescription
positiveCONDITIONING
negativeCONDITIONING
positive_textSTRING
negative_textSTRING
full_textSTRING