LoRA Loader

LoRA Loader (class AtomSimpleLoRANode) is Atom_AI's beginner-friendly multi-LoRA loader. It gives you a couple of slots by default, + / − buttons to grow them, and per-slot just three things: a LoRA dropdown, a strength slider, and an enable toggle. No clip strength, no chain, no JSON. It's the version you'd hand to someone who's never stacked a LoRA before.

The README bills it as "the simplest way to load multiple LoRAs in one node," and that's an honest pitch. Where the pack's LoRA Loader Pro exposes a separate clip strength per slot, this node deliberately doesn't - the clip strength is fixed at 1.0, and the strength slider runs 0–2 instead of the Pro node's −10 to 10. Fewer knobs, fewer ways to break the result.

How it works

Same machinery as the Pro node, minus a couple of fields. A JavaScript extension hides the node's loras_json input, draws your slots in the UI, and serializes them into that hidden JSON string. On the Python side, each enabled slot is loaded by name and fused into the model with ComfyUI's stock load_lora_for_models, with clip strength hardcoded at 1.0. Disabled slots and anything left on "None" are skipped. The node is marked always-changed, so it re-runs fresh every time rather than trusting a cache.

One design detail worth knowing: unlike the pack's stack loader, this node checks whether the LoRA file actually exists in your models/loras folder and skips it with a ✗ not found note in the info output instead of crashing. That makes it forgiving when a workflow references a LoRA you've since deleted or renamed.

Inputs and outputs

• model and clip - from your checkpoint.
• loras_json - the hidden JSON widget the +/− buttons manage. You'll never type into it.

Per slot (in the UI): a lora dropdown, a strength slider (0–2, default 1.0), and an enabled toggle.

Outputs are model and clip, wiring into your KSampler and CLIP Text Encode, plus a stack_info string that reports how many LoRAs loaded and which ones were skipped.

Installing it

It ships in the Atom_AI pack, so install once: ComfyUI Manager → search Atom_AI → install → restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/AtomAI-GR/Atom_AI

Restart ComfyUI after cloning. No requirements to satisfy and no models to download - your LoRAs just need to be in models/loras. It appears under Atom_AI ▸ LoRA.

Common issues

• Strength isn't doing what you expect. The slider caps at 2.0 and clip strength is locked at 1.0 - that's the node's whole design. If you need negative strengths or per-LoRA clip control, that's what LoRA Loader Pro is for. This node is the training wheels, not the whole bike.
• The +/− buttons are missing. The UI depends on the pack's web extension. Reinstall the pack, restart ComfyUI, and hard-refresh the browser tab; a stale tab is the usual culprit.
• "It loaded but nothing changed." A LoRA from the wrong base model family, or a missing trigger word in the prompt, is the real cause in the vast majority of cases. The community test that never fails: run the LoRA alone at strength 1.0 with a fixed seed and compare against no LoRA. If it's still inert, the file or the prompt is the problem, not the loader.

The honest take: this node exists to keep things simple, and it succeeds - for a beginner stacking two or three character/style LoRAs, it's genuinely the least intimidating option in the pack. You'll outgrow the fixed 1.0 clip strength eventually, at which point the Pro variant is the natural next step. But as an on-ramp, it's the right tool.