RM LoRA Apply

The other half of the RM LoRA system. RM Power LoRA Loader builds a stack of LoRAs as data; RMLoraApply is the node that actually does something with it. You hand it your model and CLIP plus one or more LORA_STACK inputs, and it applies every LoRA in the stack in one pass. If you've ever chained five LoraLoader nodes just to stack five LoRAs, you know exactly why this exists.

How it works

Under the hood it's not doing anything exotic - it iterates the combined stack and calls ComfyUI's own LoraLoader.load_lora() for each entry, feeding the model and CLIP through in sequence. What makes it useful is the plumbing around that loop:

• It merges every lora_stack input you connect, plus any lora_string inputs (parsed from <lora:name:weight> format), into one combined stack. You can add more inputs dynamically via "Add Lora Stack" / "Add Lora String" buttons.
• LoRAs with a strength of 0 are skipped, and LoRAs whose files can't be found are skipped with a console warning instead of aborting the run. A missing LoRA shouldn't take down a whole batch.
• The console prints a clean list of every LoRA applied with its weight - including which ones came from a random-strength range or a Civitai download.

The CLIP strength follows the model strength, which is what you want for most LoRAs (some need CLIP-skip adjustments separately, but that's a per-LoRA thing, not this node's job). If you connect no CLIP, it applies the model-only strength and passes CLIP through.

Inputs and outputs that matter

• model (MODEL, required) - from your checkpoint or UNET loader.
• clip (CLIP, optional) - connect it unless you're on a model with no text encoder in play.
• lora_stack (LORA_STACK, optional) - from RM Power LoRA Loader or RM LoRA Collector. Add more with the buttons.
• Outputs: Model and Clip - wire straight into your sampler (KSampler, or a model-patching sampler).

How to install it

cd ComfyUI/custom_nodes
git clone https://github.com/Moser9815/ComfyUI-RMAutomation

Restart ComfyUI, find it under RMAutomation/LoRA. It calls core ComfyUI code, so no extra dependencies.

Common issues

The node is deliberately forgiving, which cuts both ways. If a LoRA is silently skipped, the console warning is your only clue - the filename in your stack doesn't match anything in the loras folder, and the fix is usually a path or case mismatch (LoRA filenames are case-sensitive on Linux). Remember that order matters: LoRAs apply in stack order, and the community consensus is that bigger, more structural LoRAs should come first and finer ones after. And because this node only applies what it's given, an empty stack passes model and CLIP through unchanged - that's not a bug, that's the workflow telling you no LoRAs were requested.