EmAySee Lora Processor (9-Input Switch)
Nine LoRAs pre-loaded, one selector decides which applies
- model
- clip
- lora_object_1
- lora_object_2
- lora_object_3
- lora_object_4
- lora_object_5
- lora_object_6
- lora_object_7
- lora_object_8
- lora_object_9
- MODEL
- CLIP
A LoRA switcher that applies one of nine
EmAySee_LoraProcessor is the "which of my adapters is live right now?" node. It takes a model and clip, plus up to nine optional LORA_OBJECT inputs fed from EmAySee_LoraNameSelector nodes, and an integer lora_selector (1–9) that picks which one gets applied to the model and clip. Exactly one LoRA is active per run - it's a 9-way switch, not a stacker.
The workflow it enables is LoRA comparison without rebuilding the graph. Load nine style/character adapters into the nine slots, wire lora_selector to a seed or a random int, and batch-sweep through them to see which matches your prompt. Or drive it from a UI slider and flip through adapters live. Because only the selected slot's object is applied (empty slots just passthrough), you can leave slots unwired - you're never forced to fill all nine.
How it works
Straightforward: it grabs the object at lora_object_{lora_selector}, and if it's None - or both strengths are zero - it returns the model and clip untouched. Otherwise it applies via ComfyUI's load_lora_for_models with the shared strength_model and strength_clip. One model, one LoRA, patched in one shot.
Inputs
model(MODEL),clip(CLIP) - the base to patch.lora_selector(INT, 1–9) - which slot is active.strength_model,strength_clip(FLOAT, −10 to 10, default 1.0) - applied to whichever slot wins.lora_object_1…lora_object_9(LORA_OBJECT, all optional) - the loaded adapters fromLoraNameSelector.
Outputs: MODEL and CLIP.
Install
From ComfyUI_EmAySee_CustomNodes:
cd ComfyUI/custom_nodes
git clone https://github.com/EmAySee/ComfyUI_EmAySee_CustomNodes
# restart ComfyUI
Or via ComfyUI Manager → "ComfyUI_EmAySee_CustomNodes". No extra dependencies.
Gotchas
Two things trip people up. First, the strengths are shared across all nine slots - this is a switch, not a per-slot strength table, so if your nine LoRAs need different weights (they almost always do), you'll be tuning one pair of numbers that all nine inherit. Second, an unselected-but-connected slot does nothing, which sounds obvious until you realize the common mistake: wiring all nine to a stacker expectation and wondering why only one LoRA's effect shows up. If you actually want multiple LoRAs active at once, that's EmAySee_LoraStacker20's job - this node's job is exactly one at a time. And remember the object pipeline only: feed it LORA_OBJECTs, not string names or LORA-typed outputs.
Inputs (14)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| clip | CLIP | — | |
| lora_selector | INT | 11–9 | — |
| strength_model | FLOAT | 1.00-10–10 | — |
| strength_clip | FLOAT | 1.00-10–10 | — |
| lora_object_1opt | LORA_OBJECT | — | |
| lora_object_2opt | LORA_OBJECT | — | |
| lora_object_3opt | LORA_OBJECT | — | |
| lora_object_4opt | LORA_OBJECT | — | |
| lora_object_5opt | LORA_OBJECT | — | |
| lora_object_6opt | LORA_OBJECT | — | |
| lora_object_7opt | LORA_OBJECT | — | |
| lora_object_8opt | LORA_OBJECT | — | |
| lora_object_9opt | LORA_OBJECT | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | — |
| CLIP | CLIP | — |