Multi Selector Triple CLIP

Somewhere between "one text encoder, easy" and "four encoders, why" there's the three-CLIP case - and it's more common than you'd think. A growing batch of models pair their diffusion backbone with three separate encoder files, and stringing together three CLIPLoader nodes, a UNET loader and a VAE loader is the kind of wiring that eats an evening.

Multi Selector Triple CLIP collapses that into one node: UNET, three CLIP slots, VAE, done. Of the Sage multi-selectors, it's the one most people actually land on - the Flexible variant is nicer if you want 1–4 at runtime, but Triple is the fixed, no-fuss middle.

How it works

Same pattern as every Sage selector: this is a picker that emits model_info, a MODEL_INFO bundle of file metadata (names, hashes, Civitai info), not a loader that emits tensors. Wire model_info into Load Models or Load Models + Loras to get actual model / clip / vae outputs, or into Construct Metadata to bake an accurate metadata string into your saved PNG.

The architecture keeps selection and loading separate on purpose - the README's whole metadata story relies on the same model_info bundle flowing to both the sampler path and the metadata path so the PNG always records what you actually used.

The inputs

• clip_name_1, clip_name_2, clip_name_3 - three dropdowns listing CLIP files. That's the whole point of this variant; the slots are fixed at three, no dynamic combo to fiddle with.
• unet_name - the diffusion model file from models/diffusion_models.
• weight_dtype - default, or an fp8 option (fp8_e4m3fn, fp8_e4m3fn_fast, fp8_e5m2) for quantized loading.
• vae_name - the VAE file.

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

Installation

Same pack, same steps as the other Sage nodes. 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. The only Python dependency is dynamicprompts, and no models are downloaded by the pack.

Practical notes

If you only need one or two encoders, don't force this node - pick the Single variant or set num_of_clips on the Flexible one instead; unused slots just sit there. If you genuinely need three, mind the order: encoders are typically not interchangeable, and swapping slot 1 and slot 3 can change behavior on models with a primary/secondary encoder split.

Last thing, and it's the same caveat that follows this whole pack: it's the author's personal node set with a very small community footprint. The selectors are simple enough to be trustworthy, but you're not going to find much in the way of tutorials or bug reports beyond the repo's own examples. Start from example_workflows/ if you want a known-good graph to copy.