GALIAIS-Nodes AI Character Detail Expander

Every character prompt has the same hole: you nail the identity and the pose, and the face comes out generic because you didn't bother describing it. The AI Character Detail Expander exists to close that gap. You give it a character prompt plus a focus list, and it asks your LLM to expand the character with concrete visual details - actual haircut, actual expression, what the clothes are made of - while staying inside the areas you pointed it at.

Its name says exactly what it does, which is rarer in this pack than you'd hope. The inputs: AI服务商 (the provider object - required, it's a live API call), 角色基础提示词 (character base prompt), and 扩写重点 (expansion focus). That last one is the interesting input: it defaults to face, hair, body, outfit, pose, a comma list that tells the model which dimensions to elaborate. Want it to only do the face and hair? Change the focus. This is the input you'll actually tune, and the tooltip-friendly design means you can leave the default and still get something usable.

Outputs

• 角色细节扩写 (character detail expansion) - the fleshed-out character description, meant to feed back into your positive prompt or a character section.
• 原始响应JSON - the raw model response, for the audit trail and for debugging when the output goes weird.

Where it fits and where it burns you

This slots into the AI-refinement branch of the pack, after you've got a base character prompt from the character nodes (01–10) and want a richer render without hand-writing every eyelash. It's the most beginner-friendly of the pack's AI nodes because the default focus list is sensible.

Two honest warnings. First, it's an API node: tokens on every run, and it needs the AI Provider configured. Second - and this is the real trap - "expand a character prompt with concrete visual details" is the broadest task in this pack's AI family, and broad tasks are where LLMs add details you never asked for. The model can invent a costume, change the eye color, or drift the scene. Keep the base prompt tight, and always eyeball the 扩写 output before it goes into the sampler. If you're on a pure tag-model, expansion like this can also push you into prose that model reads poorly - this node shines most when your checkpoint actually understands natural language.

Install and gotchas

Pack-wide install, once:

git clone https://github.com/GALIAIS/GALIAIS-Nodes.git ComfyUI/custom_nodes/GALIAIS-Nodes

Restart ComfyUI (or ComfyUI Manager → "GALIAIS-Nodes"). No pip dependencies, no bundled weights. Same shared gotchas as every AI node here: provider first, health check if it errors, and the pack is brand new - 0 impressions, no community writeups, so the README's flow and the JSON outputs are your docs. Start with the default 扩写重点 and a short base prompt; that's the fastest way to learn what this model-of-the-day actually does with your style.