Load Lora From ComfyOnline

Structurally, this is the same MODEL/CLIP LoRA loader as the CivitAI- and HuggingFace-sourced versions elsewhere in this pack. What's different is where the file comes from: a model ID inside ComfyOnline (comfyonline.app), a paid ComfyUI cloud platform, rather than a public marketplace. If you're not already using ComfyOnline, this specific node probably isn't the one you want - reach for LoadLoraFromCivitAI or LoadLoraFromHF instead, which pull from sources anyone can browse without an account there.

One thing worth flagging up front

Look closely at the node's own schema and there's a small documentation slip: the comfyonline_model_id field's tooltip literally reads "The ID of the model to download from CivitAI" - copy-pasted, evidently, from the sibling CivitAI node and never updated. Don't let that confuse you; the field genuinely expects a ComfyOnline ID, not a CivitAI one, regardless of what the tooltip text says. It's a small thing, but it's exactly the kind of mismatch that makes a beginner assume they're doing something wrong when they aren't.

What it's for

Same reasoning as every other loader in this pack: fetching the file as part of the run beats requiring it to already be sitting in your models/loras folder, which matters most on stateless or freshly-provisioned ComfyUI instances. This particular node only pays off if you're already inside the ComfyOnline ecosystem - the ID field has nothing to resolve against otherwise, and the pack's README doesn't document the ID format or how to find one on that platform.

The inputs and outputs that matter

• model (MODEL) and clip (CLIP) - required, the pair the LoRA gets patched into.
• comfyonline_model_id (STRING, required, default empty) - the ComfyOnline model identifier, despite what its own tooltip says (see above).
• strength_model and strength_clip (FLOAT, default 1, range -100 to 100, step 0.01) - the usual two dials for how strongly the LoRA affects the diffusion model versus the text encoder. There's no universal right value; check whatever documentation came with the LoRA rather than assuming a default like 0.7 is safe across every architecture.
• Outputs: MODEL and CLIP, wired onward exactly like any LoRA loader's outputs.

How to install it

• ComfyUI Manager - search "comfyui-model-dynamic-loader," install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/jax-explorer/comfyui-model-dynamic-loader, then restart.

No dependencies beyond the pack itself, and nothing to pre-download - the LoRA file only gets fetched when the workflow runs and the ID resolves.

Common issues & troubleshooting

The ID field doesn't fetch anything. If you're not signed into ComfyOnline or don't have models there, that's expected - there's no public fallback documented for this node, unlike its CivitAI- or HuggingFace-sourced siblings.

Confused by the tooltip mentioning CivitAI. That's the copy-paste slip described above; ignore it and use a ComfyOnline ID here.

LoRA loads but does nothing. Same universal check as any LoRA: confirm it's built for the checkpoint architecture you're actually running, and look for a trigger word requirement - both fail silently rather than erroring.

Not sure this is the node you need at all. If your actual LoRA is hosted on CivitAI or HuggingFace, it is - this node is specifically for ComfyOnline's own library, and its siblings elsewhere in this pack cover the other two sources.