ComfyUI tutorial: prompt libraries and editable LLM output

This ComfyUI tutorial explains how to reuse image prompts through Pixaroma nodes and review LLM output before passing it to image generation. It assumes familiarity with obtaining the models from earlier episodes. The speaker requires the ComfyUI and node versions shown in the version check or newer.

The prompt library stores longer descriptions under short tags. Typing @ selects a text tag, * picks randomly from a category, and # inserts a list. Lists support shuffled, random or sequential selection. The examples combine character descriptions with art styles, then join the result with text from another node. Users can change the text order and separator, and import or export library data as JSON. Deleting the entire library cannot be undone, so the speaker recommends exporting a backup.

Pause Text sits after the text generation node. Pass mode forwards the output directly; pause mode lets users regenerate or edit it before continuing. Keep mode retains the edited prompt when running again, avoiding another LLM generation. A reference-image example uses the same controls to adjust a generated description. The speaker cautions that a fixed seed does not preserve every image detail when the prompt changes.