ComfyUI Node

LC Negative

The red prompt box, decoupled from the CLIP encoder

By lonecatone23·Created 2 months ago·Updated 3 days ago· 18
LC Negative
    • negative
    negative

    Every ComfyUI workflow has a red negative-prompt box hiding on a CLIP Text Encode node. LC Negative is that box, as its own node: a multiline text field whose only job is to hold and output the negative prompt string. It sounds pointless until you realize how often the negative prompt is shared, reused, or edited in one place and read by several. Decouple the text from the encoder and you can have one negative prompt feeding multiple encoders, or a saved prompt library where the red box lives once.

    How it works

    One required input, negative (multiline, empty by default), one output, also named negative, typed STRING. Type your negative prompt, wire the output into wherever a prompt string goes - a CLIP Text Encode's negative socket, the pack's Prompt to Conditioning, or an LC Join Strings that assembles a longer negative from parts. The node's styling (it renders red to match ComfyUI's negative-prompt convention) is applied by the pack's web files, so on the canvas it reads exactly like the box you already know, just standalone.

    That's it. There's no logic, no sanitization, no transformation. It's a text source, and being boring is the point - the plumbing layer is full of nodes whose entire job is to hold a value so the graph can fan it out.

    Why you'd reach for it

    • One negative, many consumers - if you use the same quality/NSFW-blocker negative across a full pipeline, this is the single source of truth. Change it once, every encoder updates.
    • Bypassing the encoder - with model families where the negative prompt is ignored or where you're feeding a raw prompt string downstream, having the text as a plain STRING output means you control exactly where it goes and when.
    • Prompt libraries - keep your negatives in workflow templates and swap them without rebuilding nodes.

    The honest caveat

    The prompt-engineering playbook's big 2026 finding is that on LLM-encoded models the negative prompt often does almost nothing - the encoder wrapper and guidance distillation quietly kill it. So before you invest in elaborate negative pipelines, test whether your model family even respects one. On SDXL-lineage anime models it still matters; on the newest distilled Flux-class checkpoints it's often inert. This node doesn't change that - it just makes your negative easy to edit either way.

    Install

    Part of ComfyUI LC123 Nodes (MIT, 89 nodes). ComfyUI Manager → search "ComfyUI LC123 Nodes", or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/lonecatone23/ComfyUI_LC123_nodes
    

    Restart. No extra dependencies. On hosted platforms like RunningHub, request the pack and they install it for you. The pack's companion node, Positive (LCPositive), is the green box in the same family.

    CategoryLC123/conditioning

    Inputs (1)

    NameTypeDefaultDescription
    negativeSTRING

    Outputs (1)

    NameTypeDescription
    negativeSTRING