Nodes/ComfyUI_Simple_Qwen3-VL-gguf/Simple Style Selector
ComfyUI Node

Simple Style Selector

Simple Style Selector

By KLL535ยทCreated 9 months agoยทUpdated a day agoยท 85
Simple Style Selector
    • user_prompt
    • style_name
    โ—„style_presetNo changesโ–บ
    โ—„user_promptโ–บ

    You got Qwen3-VL captioning your images, and now the output is... fine. Correct. Accurate. And completely interchangeable with the caption from any other model. That's the exact moment Simple Style Selector stops being a convenience and starts being the point: it appends a style block to your user prompt so the caption comes back in the register you actually want - photorealistic shot list, Studio Ghibli mood, cyberpunk description, whatever.

    It works on the same preset-file principle as the pack's other selectors. The style_preset dropdown is populated from the _user_prompt_styles section of system_prompts_user.json, layered on top of the pack's 300+ built-in styles. Pick one and the node takes your user_prompt (a prefix, whatever you typed) and appends the selected style's text to it.

    Inputs and outputs

    • style_preset - the dropdown. Starts with No changes and Random, then the long list of styles. Random is the fun one: the node deliberately re-rolls a fresh random style on every execution, so you can loop a batch and get a different style each pass.
    • user_prompt - optional. Your own prefix, prepended to the style text.

    Outputs:

    • user_prompt - prefix + style text. This goes into the main Qwen node's user_prompt input.
    • style_name - the name of the preset that was applied, as a string. Wire it to a text logger or save node if you want to know which style a given caption used.

    Adding your own styles

    The built-ins are decent, but the real power is dropping in your own. Edit system_prompts_user.json (in the pack folder, or ComfyUI/user/SimpleQwenVL_configs/) and add a key under _user_prompt_styles:

    {
        "_user_prompt_styles": {
            "My house style": "Describe this image as if writing alt text for a product catalog: formal, factual, mention materials and lighting explicitly."
        }
    }
    

    Restart or refresh, and it appears in the dropdown. The pattern scales - keep a handful of house styles in that file and your whole captioning pipeline suddenly produces consistent, on-brand text without retyping instructions.

    Install

    It ships in the KLL535/ComfyUI_Simple_Qwen3-VL-gguf pack - the same install as the main Qwen node. If you already have that pack, this node is already there:

    cd ComfyUI/custom_nodes
    git clone https://github.com/KLL535/ComfyUI_Simple_Qwen3-VL-gguf
    

    or search ComfyUI_Simple_Qwen3-VL-gguf in ComfyUI Manager, restart, F5.

    One gotcha

    The style text gets appended to your user prompt - it's a suffix, not a replacement. If you type a very specific instruction in user_prompt and the style is wordy, the model can weight the style block heavily and wander from what you asked. If that happens, keep your prefix short and put the real instruction in the system prompt instead. Also remember it's a text node: no model loading, no VRAM, nothing to break. Just pick a style and wire the output.

    Category๐ŸŒ SimpleQwenVL

    Inputs (2)

    NameTypeDefaultDescription
    style_presetCOMBONo changes301 options: No changes, Random, ๐ŸŽจ Photorealistic, ๐ŸŽจ Cinematic Film, ๐ŸŽจ Oil Painting (Classical), ๐ŸŽจ Watercolor, +295
    user_promptoptSTRINGโ€”

    Outputs (2)

    NameTypeDescription
    user_promptSTRINGโ€”
    style_nameSTRINGโ€”