Multi Selector Quad CLIP

Some models just have a lot of text encoders. A dual CLIP setup was already the SDXL standard, then Flux needed CLIP-L and T5-XXL, and the Qwen-image and Chroma era kept the trend going. Every so often you land on a workflow that wants three or four encoder files loaded at once, and that's exactly the gap Multi Selector Quad CLIP fills: it bundles a UNET, a VAE, and four CLIP slots in a single node.

It's the most, well, quad of the Sage Utils multi-selector family. If you only ever need one or two encoders, the Single, Triple, or Flexible variants are less cluttered. But when you genuinely need four, this is the one where you don't have to fake it with chained loaders.

How it works

Like every Sage selector, this one is a picker, not a loader. You choose the files from dropdowns and it emits a model_info output - a MODEL_INFO bundle describing the UNET, the four CLIPs, and the VAE (names, hashes, Civitai data). Nothing touches VRAM yet. You feed that model_info into the pack's Load Models or Load Models + Loras node to get real model / clip / vae tensors, or into Construct Metadata if your goal is writing an accurate A1111/Civitai-style metadata string.

The description in the README's own words: selectors like this exist so you can "select model information without loading the actual models." It's a small but real ergonomic win - pick once, wire the same bundle to the sampler path and the metadata path, and they can't drift apart.

The inputs

• clip_name_1 through clip_name_4 - four dropdowns, each listing the CLIP files in models/clip. This is the whole point of the node.
• unet_name - the diffusion model file from models/diffusion_models.
• weight_dtype - default, or one of the fp8 variants (fp8_e4m3fn, fp8_e4m3fn_fast, fp8_e5m2) for quantized UNET loading.
• vae_name - the VAE to include.

The single output is model_info, typed as MODEL_INFO, with the tooltip "Combined model info bundle including UNET, CLIP, and VAE."

Installation

It's the same pack install 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 afterwards. The only dependency is dynamicprompts; there are no model files to download.

Practical notes

If you picked four CLIPs and one of them is wrong for the architecture you're running, you'll usually find out as a dimension mismatch or a blank latent, not a clean error - the loader will happily build a bundle. Match the encoder set to what the model card actually says. And a candid caveat: this pack is the author's personal node set (they've said so on Reddit, noting barely anyone else runs it), so the multi-CLIP configs are well-tested in their own workflows but not battle-tested by a big install base. Start from the pack's example workflows if you want a known-good starting point.