Load LoRA Clip Index (NH)

The standard ComfyUI LoRA loader applies one LoRA and you need a separate loader node for each one. Run a workflow with twenty LoRAs and you've got twenty nodes. Load LoRA Clip Index (NH) (class NH_LoraClipIndex) is the pack's answer: one node that holds a menu of up to 64 LoRAs and applies whichever one you index into - to both the MODEL and the CLIP, which is why it's the "Clip" variant. It's a switching loader, built for configurable workflows where you want to flip LoRAs without rebuilding the graph.

How it works

You list your LoRAs in the optional lora_1 … lora_64 dropdowns (only the first lora_count matter), and pick which to apply two ways, via select_mode:

• index - the index number (1-based) selects which LoRA slot gets applied.
• boolean - each slot has an enable_N toggle; the first enabled slot wins. If none are enabled, the node returns a blocked model (a LoRA-less passthrough), so the graph keeps running rather than erroring.

The model and clip inputs come from your checkpoint loader, strength_model / strength_clip set the blend strengths just like the core loader, and outputs are the patched model, clip, plus filename and the index that was used - so downstream nodes know what actually ran. It reuses ComfyUI's own LoRA-loading machinery, so anything in models/loras shows up in the dropdowns.

Where you'd actually use it

Any "same graph, different LoRA per run" situation: A/B testing a style LoRA against a character LoRA, queueing a batch where each run uses a different model, or keeping one tidy loader instead of a wall of LoraLoader nodes. Drive index from a seed, a slider, or Math Eval (NH), and one workflow becomes a LoRA player. The boolean mode is the nice one for parallel-enabled workflows - flip toggles instead of typing numbers.

Installing

ComfyUI Manager → search NH-Nodes → install → restart. Or:

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

Restart and refresh the browser page. No LoRA model files download for this node - it reads whatever is already in models/loras.

Gotchas

Indexing is 1-based here - index 1 is the first slot, unlike ComfyUI's 0-based list indexing, and that mismatch has confused more than one person. Also, the dropdowns populate from models/loras at node load; if you drop a new LoRA in the folder you'll need to refresh the browser page before it appears. In boolean mode, only the first enabled slot applies - later enabled slots are ignored, so don't try to stack multiple LoRAs through one of these. For stacking, you still want separate loaders.