Load LoRA Multi (Model Manager)

Most LoRA workflows end up chaining two or three LoRA loaders in a row, each with its own strength sliders and wires. Load LoRA Multi (Model Manager) is the pack's answer to that: one node, as many LoRA slots as you want, each with its own toggle and inline strength controls. This is the flashiest node in the pack, and it's the one the README is proud of - reasonably so.

What you get

The fixed sockets are minimal: model (required, from your checkpoint loader) and clip (optional). The magic is the dynamic slots. Via the frontend you can:

• Add and remove slots on the fly - a combo dropdown per slot, populated from the server's LoRA library, grouped by base model so you don't accidentally stack an SD 1.5 LoRA on an SDXL graph.
• Toggle each slot on/off independently, plus a "Toggle All" switch.
• Adjust strength inline with arrow controls per slot, no extra widget hunting.
• Right-click a slot for a context menu to move or remove it.

Under the hood each slot becomes a lora_ input that the node reads as a structured value - on/off, LoRA selection, and strengths - so the graph stays clean while the node applies them in order with ComfyUI's normal LoRA machinery. Slots that are toggled off are skipped entirely, and if no clip is connected the clip strength is forced to zero rather than erroring.

Outputs: model and clip (the fully stacked pair), plus lora_info - one combined list of every LoRA actually applied, ready to feed the Upload Image (Model Manager) node or Merge LoRA Info (Model Manager) if you're combining with a separate branch.

How it behaves

First use of each LoRA streams it down with a progress bar into ComfyUI/models/model_manager_cache/loras/; afterwards it's cached. The node keeps loaded LoRAs in memory keyed by selection, so re-runs don't re-read files. The IS_CHANGED hook tracks the connection version, so the dropdown lists refresh when the server changes.

Install

ComfyUI Manager (search comfyui-model-manager-nodes) or:

cd ComfyUI/custom_nodes
git clone https://github.com/blackpaw-studio/comfyui-model-manager-nodes
cd comfyui-model-manager-nodes
pip install -r requirements.txt

Restart, then Connect to Model Manager (URL + API key, persisted to config.yaml; env vars MODEL_MANAGER_API_URL / MODEL_MANAGER_API_KEY override). Dependencies stay light - requests and pyyaml.

Troubleshooting

• Slots won't add or the menu's missing - this is a frontend feature, so make sure you restarted ComfyUI after the install and that your browser isn't serving a cached old frontend (hard refresh usually fixes it).
• (not connected) in the slot dropdowns - same fix as everywhere: authenticate first, and double-check the key if it keeps failing (401s kill the list).
• A stacked LoRA isn't taking effect - check that slot's toggle and strength; a toggled-off or zero-strength slot does nothing by design.
• Base-model mismatch - the grouping is there to stop this, but trust but verify: a LoRA trained for one family won't behave on another.

If your LoRA usage is "one or two per image," the single loader is simpler. If you're the person who has a permanent stack of three to five, this node will feel like it was built for you - because it basically was.