JLC - LoRA Loader Stack (10) [Legacy]

The node's own display name ends in [Legacy], which tells you most of the story. This is a fixed ten-slot MODEL+CLIP LoRA stack from an earlier version of the JLC pack, kept around as a compatibility wrapper so saved workflows don't explode when you reopen them. You are not supposed to add it to new graphs. You are supposed to keep it around so the old graphs you already have keep working.

Here's how you got here, probably: you dragged in a shared workflow, ComfyUI said "missing node type: JLC_LoraLoaderTenStack," and you searched for it. That's exactly the job this node exists for. Load the pack, restart, and the workflow opens with ten LoRA slots staring back at you.

What it actually does

It's a straightforward MODEL+CLIP LoRA stack: model in, clip in, ten LoRA dropdowns, ten strength sliders, model and clip out. Wire the model and clip outputs into your sampler and checkpoint/CLIP loader downstream, and every active slot gets applied sequentially, in slot order, before sampling.

Under the hood it's no longer its own loader. The pack rewrote the LoRA family as dynamic loaders with a shared core, and this legacy node is a thin wrapper that delegates to that core - comfy.sd.load_lora_for_models handles the actual patching. The inputs are the only part of it that's still "old."

The inputs that matter

• lora_01 through lora_10 - dropdowns populated with your models/loras folder (plus None).
• strength_01 through strength_10 - one float per slot, −10 to 10, default 0. Slots with None or zero strength are skipped, which is how you leave gaps in the stack.

And here's the legacy limitation to know about: each slot has one strength, and it's applied to both the MODEL and the CLIP/text encoder. The new dynamic loaders split those into strength_model and strength_clip so you can, say, load a style LoRA's model weights hard while leaving the text encoder alone. This node can't do that.

The outputs are simply model and clip - both patched by whatever slots were active.

Replace it if you can

If you're building anything new, use the dynamic family instead: JLC LoRA Loader - Multi Model for MODEL-only stacks, or JLC LoRA Loader - Multi-Model / CLIP when you want independent MODEL and CLIP strengths. They predeclare up to ten slots but hide inactive rows behind a slot_count control, which is a lot easier to read on a canvas than ten always-visible rows. The author's own tooltip says this node "will be deprecated in a future JLC node release," so migrating old workflows now saves you from a broken workflow later.

One piece of advice that applies regardless of which stack you use: with multiple LoRAs stacked, order matters, and it's generally worth putting your larger LoRAs first in the chain - the interaction between stacked LoRAs is where quality usually goes to die.

Install and troubleshooting

This node ships in the jlc-comfyui-nodes pack by Damkohler (J. L. Córdova), which is on the ComfyUI Registry and in ComfyUI Manager:

# ComfyUI Manager:  Install Custom Nodes → search "JLC ComfyUI Nodes"
# or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/Damkohler/jlc-comfyui-nodes.git

Restart ComfyUI after installing (or after a git pull to update). The pack has no Python dependencies beyond what ComfyUI already ships and downloads no model files - it's a lightweight install, and this node needs nothing extra.

If a workflow still reports the node as missing after install, make sure the pack actually loaded (check the terminal for errors on startup) and that you restarted rather than just reloading the page. And if you only hit this page because a workflow depends on the node: it works, it's just not where the pack is heading.