abyz22_ImpactWildcardEncode_GetPrompt

This is the little sibling of abyz22_ImpactWildcardEncode: same wildcard and LoRA machinery, minus the batch. You feed it a template with __wildcards__ and <lora:name:weight> tags, it rolls one populated prompt, loads any LoRAs, encodes it to a single CONDITIONING, and hands you the resulting text. Where the batch version is for grid workflows, this one is for a single generation that still wants variety - or for when you only have one KSampler input to feed.

Both live in abyz22/image_control, a fork of Impact Pack, and both call Impact's ImpactWildcardProcessor internally, so the dependency story is identical: Impact Pack must be installed.

Inputs and outputs

• wildcard_text - the template (multiline).
• mode - Populate rolls the wildcards, Fixed passes text through as-is.
• Select to add LoRA / Select to add Wildcard - dropdowns for clicking tags into the text.
• seed - the roll seed. Unlike the batch node there's no batch_size, so the seed maps 1:1 to one prompt.

Outputs are model and clip (post-LoRA, if any tag was applied), conditioning (a single CONDITIONING, note: not CONDITIONINGS like its sibling), and populated_text (STRING).

How it works

It defers to ImpactWildcardProcessor.process for the wildcard population, then a shared process_with_loras helper scans the text for <lora:...> tags, loads each via ComfyUI's LoraLoader, and finally runs the plain CLIPTextEncode path to build the conditioning. In other words the node does three jobs in one pass - populate, load LoRA, encode - that most workflows would string three nodes together for.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/abyz22/image_control

or ComfyUI Manager → "image_control", restart, and install Impact Pack alongside. Pack requirements add openpyxl, pytorch-lightning, kornia.

Which one to pick

If your workflow has a single KSampler and you just want a bit of prompt roulette, grab this one - one CONDITIONING out, drop it straight into the sampler's positive input. If you're building a batch/variation grid, the abyz22_ImpactWildcardEncode variant gives you CONDITIONINGS for free. Same gotchas apply here: a __missing__ wildcard passes through literally, so check the populated_text output when your images come back with angle-bracket gibberish, and point the pack's custom_wildcards folder at a single shared path if you run both this pack and Impact Pack so their config files don't disagree.