Multi Selector Single CLIP

A lot of current models use exactly one text encoder - Lumina 2, Qwen-Image, the Chroma line, and others all run with a single CLIP or LLM-style encoder file. You'd think that's the easy case, and then you stare at the CLIPLoader type dropdown and wonder whether it's lumina2, qwen_image, chroma, or one of the dozen other loader variants the core node offers.

Multi Selector Single CLIP is Sage Utils' answer: a UNET, one CLIP, and a VAE, all chosen in one node, with the loader type handled for you. It's the "I know what I want, I just don't want to remember the type string" node.

How it works

Like the rest of the selector family, this node doesn't load anything. It assembles a model_info bundle - file names, hashes, Civitai metadata - and hands it out as a MODEL_INFO output. The actual loading happens when you connect that bundle to a loader node in the same pack: Load Models (Sage_LoadModelFromInfo) turns it into model / clip / vae outputs for your KSampler, and Construct Metadata turns it into the metadata string that lands in the saved PNG.

The one input that does the real work behind the scenes is clip_type. It defaults to chroma but lists the full set of single-encoder loader options - stable_diffusion, sd3, lumina2, qwen_image, wan, hidream, cosmos, flux2, and more. If your model's encoder isn't behaving, the fix is usually this dropdown, not the file itself.

The inputs

• unet_name - the diffusion model file.
• weight_dtype - default, fp8_e4m3fn, fp8_e4m3fn_fast, or fp8_e5m2 for the UNET weights.
• clip_name - the single text encoder file.
• clip_type - the loader variant, as above.
• vae_name - the VAE.

Output: model_info (MODEL_INFO), "Combined model info bundle including UNET, CLIP, and VAE."

Installation

Standard Sage Utils install - ComfyUI Manager, search "Sage Utils", or:

cd ComfyUI/custom_nodes
git clone https://github.com/arcum42/ComfyUI_SageUtils
cd ComfyUI_SageUtils
pip install -r requirements.txt

Restart ComfyUI. Only dependency is dynamicprompts; no models to fetch.

Where people get burned

Because the node pre-fills clip_type with chroma, it's easy to leave it there out of habit and wonder why a non-Chroma model mangles your prompts. Treat clip_type as part of your model selection, not a default you can ignore. Also remember the selector/loader split: if a KSampler input is empty, check that you actually connected model_info into a Load Models node - the dropdowns alone generate nothing.

One honest caveat: this pack is a personal node set with a tiny install base, so while the selectors are simple and reliable, you won't find a big body of community troubleshooting if something odd shows up. That's a reason to keep your workflow minimal and test the bundle early.