MultipleLoraLoader10 🍌

Stacking LoRAs by chaining loader nodes works, but past three or four it turns your graph into a wall of near-identical boxes. This node collapses that into one: ten LoRA slots, each with its own name, strength, and on/off toggle, all in a single node. The pack's own description is blunt about the tradeoff - it's a fixed 10-slot loader, with the slot count set in a config.txt file in the pack rather than something you adjust per-node.

Why the apply toggles matter

Every slot has its own apply_N boolean alongside the name and strength. That's not just convenience - LoRA stacking is genuinely unpredictable (interactions between LoRAs vary, and the standard advice is to add one at a time with a fixed seed to see what each one actually does to your image). The toggles let you do exactly that without disconnecting nodes or clearing fields: load all ten, then flip them on one at a time to isolate what each is contributing before committing to the full stack.

Inputs and outputs that matter

• model - the model to patch.
• normalize (default off) and normalize_sum (default 1, range −50 to 50) - when normalize is on, the active slots' strengths get rescaled so they sum to normalize_sum instead of applying at their raw values. This is the fix for the common "stacked five LoRAs and now everything's overcooked" problem: rather than manually dividing each strength by five, let the node rebalance them for you.
• lora_name_0 … lora_name_9, strength_model_0 … strength_model_9 (default 1, −20 to 20 each), apply_0 … apply_9 (default on) - ten repeated groups of the same three fields, one group per slot.
• clip_optional - optional CLIP input if you want the LoRAs' text-encoder weights applied too.
• Outputs: MODEL and CLIP, patched with whichever slots are toggled on.

Installing it

ComfyUI Manager → search cgem156-ComfyUI🍌 → install → restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/laksjdjf/cgem156-ComfyUI

then restart ComfyUI. Nothing extra to download for the node itself, beyond whatever LoRA files you're loading into it.

Common issues & troubleshooting

Need more than 10 LoRAs. The slot count is fixed by the pack's config.txt, not by anything in the node's own UI - if ten genuinely isn't enough, that's a pack-level setting to look at, not a node option to bump.

Result looks blown out with several slots active. Turn normalize on and set normalize_sum to something reasonable (1 is a fine starting point) rather than manually eyeballing five different strength sliders down. That's precisely the problem this switch exists to solve.

Can't tell which LoRA is causing an issue. Use the apply_N toggles to bisect: turn everything off, then re-enable slots one at a time (or in halves) until the problem reappears, rather than deleting and re-adding LoRA names to test.

Empty slots doing nothing. A slot with no lora_name selected and apply on is presumably a no-op rather than an error, but leave unused slots' apply off anyway to keep the graph's intent clear for your future self.