🃏 Wildcard Processor:
WildcardProcessor prepares prompts that use wildcard syntax, helping you generate variations without manually rewriting each prompt. It passes prompt text to a wildcard loader for expansion, or lets you keep an already populated version unchanged. It also applies a selectable model preset that adds positive prompt terms and supplies a matching negative prompt. This makes it useful when you want to explore variations, preserve a particular result, or keep prompt formatting consistent across a workflow. The exact wildcard syntax and expansion choices depend on the wildcard loader and are not defined in this node’s source.
🃏 Wildcard Processor Input Parameters:
text
The prompt to process in populate mode. Enter your main prompt and any wildcard syntax supported by your installed wildcard loader here; the selected model’s positive prefix is added afterward. In fixed mode, this field is not used to form the result, and in reproduce mode the node processes populated_text instead. This is a required multiline string with a default of an empty string; no length limit is specified in the node interface.
seed
Controls the number supplied to the wildcard loader for randomization in populate and reproduce modes. Reusing a nonzero seed can help you request the same random sequence, although results may also depend on wildcard data and session-use tracking. A value of 0 asks the node to generate a random seed for that run. This optional integer defaults to 0 and accepts values from 0 through 18446744073709551615. It has no effect in fixed mode because that mode does not process wildcards.
populated_text
Holds the alternate prompt text used by fixed and reproduce modes. In fixed mode, its contents are used as-is before the model preset is applied, allowing you to edit or preserve a chosen prompt. In reproduce mode, its contents are sent through the wildcard loader rather than used literally. The interface describes it as a display of the expanded result in populate mode, but the node’s processing code reads text in that mode and does not itself assign a new value to populated_text. This optional multiline string defaults to an empty string; no length limit is specified.
mode
Chooses where the prompt comes from and whether wildcard processing runs. The options are populate (default), which processes text; fixed, which uses populated_text without wildcard processing; and reproduce, which processes populated_text. The interface describes reproduce as reusing populated text once before switching back to populate, but the provided processing code does not perform that mode switch itself. A model preset is applied to the resulting text in all three modes.
deduplicate
Controls whether the node supplies session-use tracking to the wildcard loader to reduce the probability of options that have already been used. This is an optional Boolean setting that defaults to True. Tracking is supplied only when deduplicate is True and downvote_factor is below 1.0; the loader’s exact selection behavior is not shown in the provided source. This setting does not affect fixed mode’s unprocessed text.
downvote_factor
Sets the multiplier intended to reduce the probability of an option for each previous use when session-use tracking is active. Lower values request a stronger reduction; 1.0 disables the tracking supplied by this node even if deduplicate is enabled. This optional float defaults to 0.5, has a minimum of 0.01 and a maximum of 1.0, and uses a 0.05 interface step. It is passed to the wildcard loader in populate and reproduce modes; the loader implementation determines its exact effect.
base_model
Selects the preset applied after the prompt text is chosen or processed. The options are None, Pony (default), and Illustrious. None adds no positive prefix and returns an empty negative prompt. Pony adds score_9, score_8_up, score_7_up and supplies its preset negative prompt; Illustrious adds masterwork, masterpiece, best quality, detailed, high detail, very aesthetic and supplies its own preset negative prompt. Before adding the selected prefix, the node removes a matching preset prefix already at the beginning of the text, helping avoid duplicate preset terms.
🃏 Wildcard Processor Output Parameters:
processed_text
The final positive prompt string. It contains the text selected by the mode, after wildcard-loader processing where applicable, followed by application of the selected model preset. If that preset has a positive prefix, the prefix appears before your prompt content, separated by a comma when both are present. Use this output as the positive prompt for a later node that accepts a string.
negative
The negative prompt string associated with base_model. It contains the preset’s negative terms rather than a wildcard-processed version of your prompt. With None, it is an empty string; with Pony or Illustrious, it contains that preset’s predefined terms. Connect it to a later node that accepts negative prompt text if you want to use the preset.
🃏 Wildcard Processor Usage Tips:
- Use
populateto request wildcard expansion fromtexton each run. Usefixedwith an editedpopulated_textwhen you want to keep its wording unchanged by wildcard processing. - Set a nonzero
seedwhen comparing wildcard results. For more consistent comparisons, also keep the wildcard data and deduplication settings unchanged; session-use tracking may otherwise affect selections. - Keep
deduplicateenabled with adownvote_factorbelow1.0when you want previously used options to become less likely. Set the factor to1.0or disablededuplicatewhen you do not want this node to supply session-use tracking. - Choose
Noneif you want no preset prompt terms. Otherwise, check that thePonyorIllustriouspreset fits the model and prompting style you intend to use. - If you choose
reproduce, check the resultingprocessed_textrather than assumingpopulated_textwill be returned unchanged: this mode sends that field through the wildcard loader.
🃏 Wildcard Processor Common Errors and Solutions:
WildcardLoader.load() or WildcardLoader.process(...) failure
- Explanation: The node calls the wildcard loader before producing its outputs, including in fixed mode. The loader implementation and its exact error messages are not provided, so a failure there may appear under a more specific message in your installation.
- Solution: Read the full error shown by your workflow, check that the extension and its wildcard resources are installed correctly, and confirm that the wildcard text follows the syntax supported by your loader.
processed_text is blank or contains only a preset prefix
- Explanation: In populate mode, the node starts from
text; in fixed or reproduce mode, it starts frompopulated_text. If the active field is empty, there may be no prompt content to add after the preset. ChoosingNonealso removes the preset prefix. - Solution: Check your selected
modeand fill in the field that mode uses. Select a model preset if you also want its positive prefix.
negative is empty
- Explanation: The
Nonepreset deliberately supplies an empty negative prompt. The node does not build one from your prompt text. - Solution: Select
PonyorIllustriousif you want this node to provide a predefined negative prompt, and verify that the preset suits your workflow.
Wildcards appear unchanged in processed_text
- Explanation: Fixed mode uses
populated_textwithout passing it to the wildcard processor. In other modes, whether a particular expression expands depends on the wildcard loader, whose syntax is not defined in the provided source. - Solution: Choose
populateto processtextorreproduceto processpopulated_text. If expressions still remain, check your loader’s supported syntax and available wildcard resources.
Repeated choices despite deduplicate being enabled
- Explanation: The setting is intended to reduce the probability of previously used options, not guarantee that every choice is unique. Session-use tracking is not supplied when
downvote_factoris1.0. - Solution: Keep
deduplicateenabled and setdownvote_factorbelow1.0. If repeats continue, remember that the loader controls the actual selections.
reproduce does not switch back to populate
- Explanation: Although the interface tooltip describes a one-time reproduce mode, the provided node processing code does not change the mode setting.
- Solution: Switch
modeback topopulateyourself after the run if your interface has not done so.
