ComfyUI Node

CLIP Selector

Pick your text encoder without digging through loader dropdowns

By Stibo·Created 5 months ago·Updated 2 months ago· 9
CLIP Selector
    • clip
    • is_gguf
    clip_name

    Loaders have their own model dropdowns, and that's exactly the problem: every time you want to swap the CLIP/text encoder, you open the loader, scroll the list, and hope you picked the same one everywhere. CLIP Selector is a Nifty pattern that pulls the selection out of the loader entirely - a standalone dropdown that outputs the filename as a string, which you then wire into the loader's input.

    What it is

    A single required widget, clip_name, listing CLIP / text encoder files from your clip / text_encoders folders (it reads both). Two outputs:

    • clip - the selected filename as a string (wildcard * type), wired into the clip_name input of a CLIP Loader or DualCLIPLoader.
    • is_gguf - a boolean that's true when the selected file ends in .gguf. This one earns its keep: GGUF quantized text encoders need a GGUF-aware loader, and this output lets you route to the right one automatically - or just see at a glance whether you're about to feed a quant file to a loader that can't handle it.

    Why the selector pattern is worth it

    Swap the encoder in one place and every loader wired to it changes together. No more opening three loaders to change a model. And because the selector is just a string, you can drive it programmatically - a combo switch feeding the selector, or a different selector per configuration. It's a small quality-of-life idea, but in a workflow that swaps between a heavy 2026-era LLM text encoder and a small CLIP for quick drafts, it's the difference between a two-click change and a hunt.

    One honest caveat: this node outputs a name string. It doesn't load the model. You still need the actual CLIP loader to do the loading - the selector just decides which file. That division is deliberate (the README frames selectors as "keep your actual loader nodes clean"), but it's a thing to know before you wire this and wonder where the model went.

    Install

    Part of Nifty Nodes for ComfyUI. Manager: search "Nifty Nodes for ComfyUI" → Install → restart. Manual:

    cd ComfyUI/custom_nodes
    git clone https://github.com/Stibo/comfyui-nifty-nodes
    

    Restart ComfyUI. No models, no downloads - it just lists files you already have. The pack is MIT-licensed, v3-API native, and new (v2.1.0, mid-2026).

    Gotchas

    If the dropdown shows [no models found], your files aren't in ComfyUI/models/clip or text_encoders (or the model is an extension the selector doesn't list - it covers .safetensors, .gguf, .pt, .pth, .bin). And remember: a .gguf text encoder needs the GGUF loader; wiring its name into a plain CLIP loader will fail. That's precisely why the is_gguf output exists - use it.

    Categorynifty/selectors

    Inputs (1)

    NameTypeDefaultDescription
    clip_nameCOMBOSelect a CLIP / text encoder model from the clip / text_encoders folder.

    Outputs (2)

    NameTypeDescription
    clip*
    is_ggufBOOLEANTrue if the selected model is a GGUF quantized file. Useful for routing to a GGUF-compatible loader.