Anima LoRA Preset

The Anima workflow has a hidden tax: a "style" isn't one LoRA, it's a LoRA stack plus a style prompt plus a pile of trigger words, and you rebuild it by hand every time you switch looks. Anima LoRA Preset is the pack's answer - a multi-profile node that stores up to 16 LoRA presets, each with its own style prompt, LoRA stack, and trigger words, then hands the whole thing to the graph as a LORA_STACK when you need it.

What it is

The node holds multiple profiles managed through front-end buttons. Each profile has a style_prompt (artist tags, model triggers, or short style directions), a list of LoRAs with their weights, and the trigger words those LoRAs need. You flip profile_index and the node swaps the entire setup.

The inputs look odd at first because most of them are internal: profile_count, loras and profile_data are serialized JSON the front-end manages, and lora_name is a hidden LoRA-selector source. The one you'd wire externally is profile_index - it can be connected from another node, so you can rotate styles programmatically (feed it a seed-driven INT, for example).

The outputs that matter

• LORA_STACK - the built stack, compatible with LoRA stack loaders and the AiO Generator's lora_stack socket. This is the one that does the work.
• style_prompt - the corrected style prompt for the selected profile.
• trigger_words - trigger words collected from the selected LoRAs' metadata, ready to paste into a prompt editor or a trigger field.
• active_loras - a text representation of the enabled LoRAs and strengths, handy for logging or display.
• profile_index - the selected index after wrapping to the available profile count.

Wire LORA_STACK into a stack loader (or the AiO Generator), wire style_prompt and trigger_words into your prompt editor, and a profile switch changes both the model patches and the prompt in one move.

How profile data survives

The nice part: profile data is preserved in workflow metadata. Save the workflow and the profiles come with it - reloading on another machine keeps your stacks even if the LoRA files differ. The pack also recommends ComfyUI-Lora-Manager alongside this for trigger-word and metadata management, since the trigger words here come from LoRA metadata and Lora-Manager is the standard way to keep that metadata clean.

Install and gotchas

Pack install:

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

Restart ComfyUI; Manager: "ComfyUI EasyUse Anima". Preset profiles are stored in ComfyUI user data, so they survive Manager updates and re-installs.

The gotchas: trigger words only populate if the LoRA files actually carry trigger metadata - if a LoRA comes from a bare safetensors with no metadata block, trigger_words comes back empty and you'll need to type them or add them via Lora-Manager. And the style_prompt output is the corrected style prompt - it runs through the pack's ANIMA normalization, so if it comes back reordered, that's intentional. Don't hand-edit loras or profile_data; use the profile buttons, or you'll desync the front-end from the stored JSON.