Load Clip Index (NH)

Every text encoder has its own CLIP model, and if you run multiple model families you end up with a lot of .safetensors files in models/clip and a lot of scrolling through dropdowns. Load Clip Index (NH) turns that dropdown into a number: it lists up to 64 CLIP files in one node, and you pick one by a 1-based index. Hand the index to a Counter (NH), a random node, or a math node, and the loader swaps CLIPs automatically between runs.

It's the same idea as the pack's indexed LoRA and diffusion loaders - one big config surface, one wire that chooses. You pay a little setup time up front; you save a lot of manual swapping later.

How it works

The node reads every file in ComfyUI's models/clip folder and lays them out as clip_1 through clip_64 inputs. Two controls shape the selection:

• select_mode - index picks by the index widget; boolean ignores the number and picks the first enable_N checkbox you've ticked. Boolean mode is the lazy person's toggle: flip the enable box on the CLIP you want this run, and the loader follows.
• clip_count - how many of the 64 slots are active (so you only see the ones you've actually populated). Set it to your real count and the rest disappear.
• type - which CLIP architecture to load: stable_diffusion, stable_cascade, sd3, stable_audio, mochi, ltxv, pixart, or hunyuan_video. Get this wrong and the load fails or produces garbage, so match it to the checkpoint family you're pairing with.

The loader uses ComfyUI's standard load_clip path with your models/embeddings folder attached, so your text encoders behave exactly like the built-in Load CLIP node.

Outputs

• clip - the loaded CLIP, ready to wire into a CLIP Text Encode node.
• filename - the name of the file that got loaded. Great for logging or for a Save Image caption showing which CLIP produced the result.
• index - the index that was selected, passed back out so downstream nodes know what happened.

If the selected slot is empty or out of range, you get a blocked output instead of a crash - downstream text encoding simply doesn't run. That's the safe-fail behavior, and it's usually what you want while you're still populating slots.

Installing

cd ComfyUI/custom_nodes
git clone https://github.com/jetthuangai/NH-Nodes.git
cd NH-Nodes
pip install -r requirements.txt

Or install NH-Nodes via ComfyUI Manager and restart. No model downloads - the node just reads whatever CLIPs you already have. First run will show empty dropdowns until you assign files to slots, which trips people up for about ten seconds.