Lora Stack Loader 10 PoP

The pack's own README calls this one "a bit ridiculous," and it's right. This is the ten-slot version of LoraStackLoader_PoP - same idea, same on/off switch per slot, same paired strength_model/strength_clip sliders, just scaled from three LoRAs up to ten. If you've ever wanted to stack a base style LoRA, a character LoRA, three detail LoRAs, and a handful of concept LoRAs all in one workflow without your graph turning into a wire jungle, this is that node. Whether you should want that is a separate question - LoRAs interact, and ten of them fighting over the same model weights is more likely to mush your output into noise than to combine cleanly. But the option's there.

How it works

Mechanically it's identical to the 3-slot loader, just looped ten times: model and clip come in, and for each of the ten slots the node checks the switch - if it's Off, that slot is skipped entirely, strength values and all, and the loop moves to the next one. Only slots switched On get their LoRA weights merged into the running model and clip. The single MODEL/CLIP pair comes out the other end carrying every enabled LoRA's contribution stacked in slot order.

The inputs and outputs that matter

• model / clip in, same as any LoRA loader.
• switch_1 through switch_10 - Off/On per slot. With ten of these on one node, get comfortable scrolling.
• lora_name_1 through lora_name_10 - pick a LoRA per active slot.
• strength_model_N / strength_clip_N, range −10 to 10 - and unlike its 3-slot sibling, this one defaults to 1, not 0. Flip a switch On here and the LoRA applies at full strength immediately, no extra step needed. (Yes, the two loaders in the same pack default differently - worth knowing if you use both.)
• MODEL and CLIP out - feed straight into your sampler and text encoders.

How to install it

ComfyUI Manager: search "comfy_PoP", install, restart. Manual:

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

Restart ComfyUI and both LoRA loaders (plus the rest of the pack - DALL-E, canny, VAE, attention, aspect ratio, and conditioning nodes) show up together. No extra dependencies, no models to download - it's a thin wrapper around ComfyUI's own LoRA machinery.

Common issues & troubleshooting

The node is enormous on your canvas. Ten slots times four fields each is forty widgets on one node. That's the actual trade-off you're making versus three separate loaders or a chain of them - collapse it or resize it to keep your graph readable, and consider whether you actually need ten active LoRAs at once before reaching for this over the 3-slot version.

Output looks muddy or over-cooked. This is a real risk with heavy LoRA stacking generally, not a bug in the node - piling on many LoRAs at once, especially at strength 1.0 each, tends to fight itself. Start with fewer switches On and build up rather than flipping everything at once.

A LoRA doesn't seem to do anything. Check its trigger word is actually in your prompt - this node has no trigger-word lookup UI, it just applies weights. If your model line is a v-pred variant (NoobAI, some Illustrious builds), be aware that stacking non-vpred LoRAs on top is a known way to degrade output rather than improve it - that's a model-compatibility issue, not something this loader can fix for you.