Character Prompt Builder - Female Poses

If you've ever typed "girl standing with hand in hair" into a prompt box and watched the model hand you back a robot arm, this node is the fix. Character Prompt Builder - Female Poses is the third link in the pack's chain (Person → Fashion → Poses → Scene & Generate), and its whole job is turning a dropdown selection into a full, grammatical pose sentence that a text encoder actually understands.

The one thing to understand before you touch it: the four pose dropdowns are mutually exclusive. The node picks the first non-"-" pose in the order standing → kneeling → sitting → laying down and silently discards the rest. So don't fill in "standing" and "sitting" hoping for variety - pick one category, leave the other three on "-", and let the Scene node handle everything else. People get burned here because the node doesn't yell at you; it just ignores your second choice.

How it works

Every node in this pack passes a PM_SETTINGS dict down the chain. The Poses node takes the settings it received via settings_in, stamps its pose data onto the dict, and hands the whole thing out the settings output. Nothing is generated yet - that happens at the end in the Scene & Generate node, which is the only node that emits text. So this node does zero prompt magic on its own; it's pure "record what you chose."

What you can record is surprisingly deep:

• Poses - standing (92 options), kneeling (16), sitting (29), laying down (17). These are written as complete phrases, e.g. "standing in an elegant S-curve with weight resting on one hip" or "twirling with dress or skirt flowing in motion."
• Props - 90 options, from "holding an iPhone 17 Pro" to "drinking from a coffee cup," plus a props_color dropdown (48 colors) that styles the prop.
• Placement (optional) - hand_placement (11), leg_placement (10), and head_placement (8) let you refine a chosen pose, like "hands on hips" or "head tilted to one side."
• custom_pose - a free-text box if none of the ~150 built-in options say what you need. It gets merged into the final prompt verbatim.

Wiring it up

Female Person -> Female Fashion -> Female Poses -> Scene & Generate -> CLIP Text Encode

Connect the settings output of the Fashion node to settings_in, and send the Poses settings output to Scene & Generate's settings1 (or settings2/3/4 for a multi-character scene). For a minimal workflow you can feed it straight from the Person node - the pack's whole selling point is that you only chain what you need.

Installing

This is the easiest install in the ComfyUI ecosystem. Through ComfyUI Manager, search "Character Prompt Builder" and hit Install. Or drop it in manually:

cd ComfyUI/custom_nodes
git clone https://github.com/euan-gwd/comfyui-character-prompt-builder

Then restart ComfyUI. There is no requirements.txt to satisfy - the pack is pure Python standard library, and there are no model files to download. First run may fetch character_prompt.json from GitHub if the shipped copy is missing, so let it sit a second if the dropdowns look thin.

Troubleshooting

• Node missing after install - restart ComfyUI fully; it's the #1 cause. If the console shows an error, check the folder landed in custom_nodes/.
• Dropdowns empty or tiny - the JSON data file failed to download and it fell back to a minimal default list. Check your connection, or re-clone so the bundled resources/character_prompt.json is present.
• My pose isn't in the image - remember only one of the four pose categories wins. Set the others to "-", and if you still get nothing, use custom_pose - free text always beats fighting the dropdowns.