🦊 Isulion Animal Behavior Generator

The Isulion Animal Behavior Generator is the pack's answer to a very specific problem: your animal prompt is great at describing what the animal is and completely silent on what it's doing. A static "fox" is a taxidermy exhibit. "Fox hunting through snow" is a picture. This node is a 40-entry behavior dropdown - Hunting, Sleeping, Playing, Stalking, Nursing, and so on - plus a randomize toggle and a seed, and it hands back the behavior as a string.

It's designed to sit alongside the pack's Animal Selector (🦁 Isulion Animal Selector), which picks the species. The intended chain: Animal Selector → Animal Behavior → your prompt. Species from one, action from the other, and suddenly "lion" becomes "lion stalking through tall grass." If you've ever gotten a beautiful animal render that's missing all life, this is the missing half.

The inputs

• behavior - 40 presets, the manual choice. Defaults give you the full spread from "Hunting" and "Stalking" (dramatic) to "Resting" and "Grooming" (calm). It's a curated list, not an infinite vocabulary, and that's fine - animals mostly do ~40 things.
• randomize - enable/disable. Enabled, it draws a random behavior; disabled, the dropdown value passes through.
• seed - INT, default 0. The pack-wide convention applies: set a seed > 0 to reproduce a specific behavior draw; leave 0 and the node rolls a fresh seed, returning the actual value on the seed output so you can lock it later.

Outputs are behavior (the phrase, as a STRING) and seed (INT). Bare phrase again - you're expected to bolt it into a larger prompt. "Hunting" on its own is not a prompt; "a {species} hunting" is.

Where it fits and where it trips people

Because it outputs a bare action word, this node is at its best in a prompt-builder graph where you concatenate species + behavior + location + style into one positive prompt. Drop it alone into a CLIP Text Encode and you'll generate a picture of... "hunting". So wire it properly: subject + behavior + environment.

The other thing to know is that behavior words and image models have opinions. Broad, physical behaviors ("running", "swimming", "flying") render reliably almost everywhere. Abstract ones ("exploring", "gathering") depend on the model's captioning and can come back muddy - a model trained on natural-language captions handles them better than a strict tag model. If a behavior isn't landing in your renders, that's the model, not the node.

Install

Same as every Isulion node: ComfyUI Manager → search "Isulion", or clone https://github.com/Isulion/ComfyUI_Isulion into custom_nodes/ and restart. No dependencies or model files - it's a hardcoded list and a random call.

It's a small node with a narrow job, and it does that job with zero ceremony. Worth grabbing if you do wildlife or pet-style generations and want the action part of the prompt to stop being a coin flip.