LoRA Selector Multi-Output

A LoRA's effect depends heavily on its strength, and the only honest way to know the right one for your checkpoint is to compare a few renders side by side. Normally that means several LoraLoader nodes, each loading the same file. LoRA Selector Multi-Output is the single-node version: pick a LoRA once, get seven MODEL outputs, each with the LoRA applied at a different strength. Wire each output to its own KSampler and run them all in one batch.

The mechanism

The node loads the LoRA file from disk exactly once, then calls ComfyUI's load_lora_for_models seven times - once per strength slot. A strength of 0 returns the original model untouched (no LoRA applied), so you can keep one output as your baseline. Everything else gets the patch at that strength.

Worth knowing: the LoRA is applied to the model only - the CLIP strength is always 0. For most character/style LoRAs that's exactly what you want; if a LoRA relies on prompt-side CLIP influence, you'll want a regular LoraLoader for that one.

Inputs and outputs

• model (MODEL) - your base checkpoint's model output.
• lora_name - dropdown populated from your models/loras folder.
• strength_1 through strength_7 - each defaults to 1.0, range −10 to +10. Set them to your sweep values: e.g. 0.5 / 0.7 / 0.8 / 0.9 / 1.0 / 1.2 / 1.5.

Outputs: MODEL_1 through MODEL_7, one per strength. Note the output names - they're not "lora_1", they're the boring but unambiguous MODEL_1..7.

When to reach for it (and when not)

Reach for it when you're tuning: pick your character LoRA, sweep strengths across seven samplers, and compare in one queue. That's the workflow it was built for, and it saves you the seven loader nodes plus the tab-switching.

Don't reach for it if you need different LoRAs on different samplers - there's one dropdown, so all seven outputs carry the same LoRA. And be aware of the memory cost: each strength output is a full MODEL patch in memory, so a 14B model sweep is seven heavyweight copies. On a big model you might want to run fewer than all seven at once by leaving strengths at 0 (which bypasses the patch entirely and reuses the original).

Installing this node

One of the 16 nodes in the lzits Nodes pack (Lior Zitser, MIT). Install via ComfyUI Manager (search "lzits Nodes") or:

cd ComfyUI/custom_nodes
git clone https://github.com/lzitser23/lzits-nodes.git

Restart ComfyUI. No extra dependencies - it uses ComfyUI's own LoRA loading internals.