Lora List Loader

The usual way to stack LoRAs in ComfyUI is to chain LoRA loaders - one node per LoRA, each with its own strength slider, daisy-chained down the graph. Fine for two. Annoying for five. Impossible to drive from an API. This node collapses the whole chain into one: it takes a text list of LoRAs, applies all of them to your model and CLIP in a single pass, and spits the modified pair back out.

The reason this exists isn't laziness, it's automation. When a workflow is being driven by something outside ComfyUI - a script, an app, an API call - you can't rewire the graph per request. But you can change a text field. So instead of a fixed chain of loader nodes, you send a string that says "these LoRAs, at these strengths," and the same workflow serves a different LoRA stack every run. That's the ArtVenture pattern: push configuration into data so the graph stays static.

How it works

You feed it a model, a CLIP, and a blob of text describing which LoRAs to load. The node parses that text into a list, then applies each LoRA to both the model and the CLIP in sequence - exactly what a stack of individual loaders would do, just done in one node from a data source instead of from wired-in dropdowns. The LoRA files themselves still have to live in your models/loras folder; the text just names them and sets strengths.

Inputs and outputs

Three required inputs:

• model - the MODEL to apply the LoRAs to, from your checkpoint loader.
• clip - the matching CLIP. LoRAs typically patch both, which is why both come in and both go out.
• data - the multiline string listing the LoRAs and their weights. This is the field that makes the node worth using; it's what an external caller overrides per request.

Two outputs, MODEL and CLIP - the patched pair, wired onward to your sampler and text encoders just like the output of any LoRA loader.

Installing it

Part of the ArtVenture pack. ComfyUI Manager: search comfyui-art-venture, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/sipherxyz/comfyui-art-venture

Restart ComfyUI. No extra model download for the node itself - but every LoRA you name in data must already be sitting in ComfyUI/models/loras, or the node has nothing to load.

Common issues

A named LoRA isn't found. The names in data have to match the actual filenames in your loras folder. Typos, wrong subfolder, missing extension - any of these and that entry silently does nothing or errors. Confirm the file exists and the name matches exactly.

Nothing changed in the output. Usually a malformed data string, so nothing parsed, or strengths set to zero. Start with a single known-good LoRA at strength 1.0 and build up - that isolates whether it's the format or the file.

Too many LoRAs, mushy image. Not a bug, just LoRA reality: stack enough of them at high strength and they fight, colors blow out, the composition degrades. Pull individual strengths down when you stack. This node makes over-stacking easy, which means it also makes it easy to overdo.

It won't take a base with a different CLIP. Model and CLIP have to be from the same checkpoint family. A LoRA trained for SDXL applied to an SD 1.5 model is a mismatch, and no loader saves you from that.