Anima Wildcard LoRA

The same wildcard node, plus a LoRA pickup line

Anima Wildcard expands __name__ tokens and {a|b|c} choices into real prompt text. This is that node - same inputs, same four modes, same seeded determinism - with one genuinely useful addition: it also recognizes <lora:...> tags in the prompt, strips them out of the text, and hands them to you as a proper LORA_STACK you can wire straight into a LoRA loader. No regex wrangling, no "my LoRA name ended up as literal text in my prompt" surprise.

It exists because on ANIMA (Circlestone's 2B anime model, which the whole EasyUse Anima pack orbits) LoRA stacking is the main way people bend the base aesthetic toward a specific character or style. If your wildcard files select both scenery and a character LoRA, this node lets one pass do both jobs.

How it works

It runs the identical wildcard engine as the base node, then post-processes the expanded output. Any <lora:name:strength> or <lora:name:strength:clip> tag (the single-bracket A1111 form) - and the double-bracket <<lora:...>> form ComfyUI-Lora-Manager uses - is pulled out of the text. If no clip strength is given, the model strength is used for both. The tag disappears from the prompt, and the node resolves it against your models/loras folder into a stack entry.

The output LORA_STACK is your optional input stack (if you fed one in) followed by whatever the prompt referenced, in prompt order. The base node's lora_stack input is optional and just gets prepended - handy when you have a "always on" style LoRA plus prompt-driven ones.

Inputs and outputs that matter

Same set as the base node: text, mode, seed, and populated_text with the same General / Fixed / Sequential / Reproduce behavior, plus one optional lora_stack input for your always-on LoRAs. Outputs are text (the expanded prompt, LoRA tags removed), seed, and LORA_STACK (the merged, ordered stack).

The workflow shape is: text → this node → text into CLIP Text Encode and LORA_STACK into a LoraLoader. Or feed LORA_STACK into the AiO Generator if you're running the pack's all-in-one flow.

Installing it

Same pack, same install as the base node - you get both at once. ComfyUI Manager (search "ComfyUI EasyUse Anima") or:

cd ComfyUI/custom_nodes
git clone https://github.com/n0va39/ComfyUI-EasyUseAnima
pip install -r ComfyUI-EasyUseAnima/requirements.txt

Restart, and the default wildcard folder ComfyUI/user/__easyuse_anima/wildcards/ is created for you. Register extra wildcard folders in Settings → EasyUse Anima → Wildcard. The only real dependency to think about: LoRA tag names are resolved against whatever's actually in your models/loras folder, so a typo in a <lora:...> tag leaves you with a missing-LoRA error down the line rather than silently nothing.

Gotchas

• Order matters - extracted LoRAs append after your input stack, in prompt order. If a wildcard references three LoRAs, that's three stack entries.
• Naming - the node handles A1111's <lora:name:1.0> and LoraManager's <<lora:name:1.0>>, but it needs the name to match the file in models/loras. LoraManager's whole point is keeping those names tidy; if your LoRA filenames are v1.2_merge_final_fixed.safetensors noise, fix the filenames or you'll be chasing resolution failures.
• Same wildcard gotchas as the base node - missing wildcard files show up in the node's status, and huge seeds can exceed browser integer precision.

It's a thin wrapper, honestly - but it's the wrapper that turns prompt-side LoRA tags into something a ComfyUI graph can actually consume, which is the part people usually end up hand-rolling. Nice to have it built in.