Lora Stack Loader PoP

If you've ever wired up three separate LoraLoader nodes to stack three LoRAs, you know the annoyance: three model/clip pairs threading across your graph, and disabling one means unplugging wires instead of flipping a switch. LoraStackLoader_PoP is the fix - three LoRA slots baked into a single node, each with its own on/off toggle and independent model/clip strengths. The pack's own README undersells it: "isn't exactly breaking new ground, but couldn't find a node that had all the attributes I was looking for." That's the whole pitch, and it's a reasonable one - this is a quality-of-life node, not a clever one.

How it works

You feed it your checkpoint's model and clip, and for each of the three slots you get a switch (Off/On), a lora_name, and separate strength_model / strength_clip sliders. The node loops through the three slots in order and applies each LoRA's weight patch to the running model/clip - but only if that slot's switch is On. Flip a switch to Off and that slot is skipped completely; the code just moves to the next one, so a stray strength value sitting in a disabled slot does nothing. That's the entire value proposition: toggle LoRAs in and out without rewiring or deleting nodes.

The inputs and outputs that matter

• model / clip in - from your checkpoint loader.
• switch_1/2/3 - Off or On per slot. This is the thing you'll actually click during iteration.
• lora_name_1/2/3 - pick from your installed LoRAs, one per slot.
• strength_model_N / strength_clip_N - range −10 to 10, and here's the trap: default is 0, not 1. Flip a switch On and forget to raise the strength, and that LoRA loads but contributes nothing, because you're multiplying its weights by zero. Most LoRA loaders you've used default to 1.0 - this one doesn't, so get in the habit of setting strength right after you flip the switch.
• MODEL and CLIP out - wire straight into your KSampler and CLIP Text Encode nodes, same as any LoRA loader chain.

How to install it

Via ComfyUI Manager: search "comfy_PoP", install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/picturesonpictures/comfy_PoP

Then restart ComfyUI. There's no requirements.txt weight to speak of here and nothing to download - this is pure Python logic wrapping ComfyUI's own LoRA-loading calls, so once the clone's in place you're done. The whole pack installs the same way; if you end up wanting the DALL-E, canny, VAE, or attention nodes from the same author later, you already have them.

Common issues & troubleshooting

Switch is On, nothing changed. This is the default-zero trap above - check the strength sliders, not just the switch.

Trigger word does nothing. Not this node's fault: most LoRAs need a specific trigger word typed into your prompt to activate, and this loader has no UI for surfacing what that word is (unlike, say, rgthree's Power Lora Loader, which can pull trigger words off CivitAI for you). Check the LoRA's model page if your output looks untouched even with strength cranked up.

Only need one or two LoRAs? Leave the unused slots' switches Off - a disabled slot costs nothing and doesn't need a valid lora_name to be harmless, since it's skipped before the name is ever touched.

Need more than three? That's the whole reason LoraStackLoader10_PoP exists in the same pack - identical inputs/outputs, just ten slots instead of three.