DP Five Lora

Stacking LoRAs the vanilla way means chaining five separate Load LoRA nodes, each threading model and clip into the next, each with its own strength field. It works, but it's a wiring headache and it clutters the graph. DP Five Lora collapses that into one node: pick up to five LoRAs, set their strengths, and pass model and clip straight through.

If you've heard people swear by rgthree's Power Lora Loader, this is the same idea from a different pack - one node that stacks multiple LoRAs with per-entry control. DP Five Lora doesn't have Power Lora Loader's trick of reading trigger words off the file or pulling them from CivitAI, so it's a lighter tool. But if all you want is "load these five and let me toggle them," it does exactly that without the extra weight.

The inputs that matter

• model (MODEL) and clip (CLIP) in - your checkpoint's model and clip. They pass through, LoRA-modified, to the outputs.
• loader_state (ON / OFF) - the master switch. OFF bypasses the whole node so your base model comes through untouched. This is genuinely useful: A/B a generation with and without the entire LoRA stack by flipping one control, no rewiring.
• Lora_01 (required) plus Lora_02–Lora_05 (optional) - your LoRA slots. In the node they're dropdowns listing your installed LoRA files. (If a slot shows only "None", you've got no LoRAs in your ComfyUI/models/loras folder yet - drop some in and they'll populate.)
• strength_control (optional) - a Desert Pixel widget (LORA_STRENGTHS) that holds the per-LoRA strength values so you can dial each one independently.

Outputs: model and clip (wire these onward to your sampler and text encode) plus lora_info (a string summarizing what got loaded - route it to a text preview to confirm the stack).

How it works

Each LoRA is a small weight patch trained on top of a base model. Loading several applies them in sequence, and strength scales how hard each one pushes. The catch beginners hit is that LoRAs interact - three at 1.0 each will fight, and the loudest one wins. Balancing strengths (drop the ones that dominate, keep the ones you want subtle) is the real skill, and having all five strengths visible in one node makes that balancing act far easier than hunting across a chain.

How to install it

ComfyUI Manager: search ComfyUI-Desert-Pixel-Nodes, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/DesertPixelAi/ComfyUI-Desert-Pixel-Nodes

then restart. The node itself needs no download - but it only lists LoRA files you've already placed in ComfyUI/models/loras. Find it under the "DP" loaders category.

Common issues & troubleshooting

The dropdowns are empty / only show "None". No LoRA files found. Put .safetensors LoRAs in ComfyUI/models/loras and refresh (restart ComfyUI or hit refresh in the menu).

A LoRA does nothing, or wrecks the image. Almost always an architecture mismatch - an SD 1.5 LoRA on an SDXL checkpoint (or vice versa), or a Flux LoRA on SD. LoRAs are base-specific. Check the LoRA's model family matches your loaded checkpoint. If it matches and still misbehaves, its strength is too high or it's colliding with another in the stack - pull it down and see who was fighting.

The whole stack feels muddy. Too many LoRAs at full strength. Flip loader_state to OFF to confirm the base is clean, then re-enable and bring strengths down one at a time. Fewer LoRAs at balanced strengths beats five all shouting at 1.0.