🍭大炮-Qwen3VL额外选项@炮老师的小课堂

Qwen3VL_ExtraOptions is not a node you run by itself. It's a switchboard: seventeen booleans that get appended as extra instructions to whatever captioning prompt the other nodes in this pack are using. It exists because the author got tired of retyping the same rubric - "include lighting details, describe composition, don't mention text" - every time. Plug it into the 🎯 Qwen3VL额外选项 socket on Qwen3VL_Advanced, Qwen3VL_Batch_Caption, or Qwen3VL_Chat, flip a few toggles, and every caption those nodes produce suddenly carries your photography obsessions with it.

How it works

Mechanically it's the simplest node in the pack. Each toggle maps to one instruction sentence; when you run the workflow, the target node calls build_enhanced_prompt() and appends your enabled instructions to the base caption prompt as a "please follow these additional requirements" block. No toggles enabled means the base prompt passes through untouched - the node becomes a no-op, so there's no downside to leaving it wired in.

The toggles, grouped by what they're for

• Subject and character control: 👤 包含人物信息 (include names/roles if people are present), 🚫 排除不可改变特征 (skip race/gender, keep changeable stuff like hairstyle - a genuinely thoughtful default for training captions, since fixed traits are exactly what you don't want baked into a character LoRA).
• Photography nerdery: 💡 包含光照信息, 💡 提及光源, 📐 包含相机角度, 📷 包含相机详情 (aperture/shutter/ISO), 📊 包含构图信息 (rule of thirds, leading lines), 🌈 包含景深信息.
• Quality assessment: 🎨 包含艺术质量 (aesthetic quality on a low-to-high scale), ⭐ 描述重要元素.
• Content filtering: 🔍 排除性感内容, 🔒 包含安全性, 📝 不提及文字, 🔇 不提及分辨率.
• Technical honesty: 🏷️ 包含水印信息, 🖼️ 包含JPEG伪影 (compression artifacts), 🌍 不使用模糊语言 (no vague wording).

The names are self-describing in Chinese, which is a small speed bump if you don't read it - the tooltips explain each one, and the English meanings above cover the full set. 包含人物信息 + 🚫 排除不可改变特征 + 💡 包含光照信息 + 📐 包含相机角度 is a strong default combo for turning plain captions into photoreal-style training captions; flip them all off and you're back to stock behavior.

Installing

It installs with the pack - ComfyUI Manager (search ComfyUI-Qwen3VL-DP) or:

cd ComfyUI/custom_nodes
git clone https://github.com/paolaoshi/ComfyUI-Qwen3VL-DP
cd ComfyUI-Qwen3VL-DP
pip install -r requirements.txt

then restart. There's no model download of its own - the heavy lifting (and the transformers >= 4.57.0 requirement, and the model downloads) all belongs to the node you attach it to. If you hit an "unable to import Qwen3VL extra options module" error in older versions, update the pack: the author fixed that import bug and the current release handles it cleanly. If the extra options don't seem to be doing anything, check the console for the "已应用Qwen3VL额外选项增强提示词" line (or the count of enabled options) - that confirms the socket actually got the connection.