Draw Things Pipeline Add Lora

This little node does one job: it bolts a LoRA onto a Draw Things pipeline mid-graph. It sits in the "advanced" flow between a Txt2Img/Img2Img Pipeline node and the DrawThingsGenerateFromPipeline node, and it edits the settings blob on its way through to add your adapter and its weight.

Here's the thing worth knowing: you don't strictly need this. In basic mode, whatever LoRAs you've got loaded in the Draw Things app already apply - the wrapper respects the app's settings unless you override them. What this node buys you is control from inside ComfyUI: you set the LoRA and its strength on the graph instead of clicking around the app, which matters when you're building a repeatable workflow or want to sweep weights across a batch. If you're happy setting LoRAs in Draw Things by hand, you can skip this node entirely.

A LoRA, if you're new to the term, is a small low-rank adapter that patches a checkpoint to add a character, style, or concept without retraining the whole model. The catch everyone hits once: LoRAs are architecture-bound. A LoRA trained for SDXL does nothing on Flux. Whatever LoRA you name here has to match the base model in your pipeline.

How it works

The node takes the incoming pipeline dict, injects your LoRA reference and weight into the Draw Things config, and passes the modified dict along. Nothing generates here - it's a pass-through editor. The render only happens when the dict reaches DrawThingsGenerateFromPipeline and gets POSTed to the Draw Things API.

The inputs and outputs that matter

• pipeline (dict) - the config coming from a pipeline node upstream. Required.
• lora (string) - the LoRA name, and this is the gotcha: it must match the Draw Things LoRA filename exactly. Best way to get it right is Draw Things' "..." → Copy configuration, which prints the exact model and LoRA names; paste that and copy verbatim.
• weight (default 1.0, range 0–2.5) - strength. The old "0.5–0.8" rule of thumb is SDXL-era advice; on Flux and newer bases people often run 1.0 and up because dropping it costs likeness. Start at 1.0 and adjust to taste.

Output is pipeline (dict) - chain into another Add-node to stack a second LoRA or a control, or into DrawThingsGenerateFromPipeline to run.

How to install it

Comes with the Draw Things pack. In Draw Things, open Advanced → enable API Server (HTTP, Port 7860, localhost). Then install via ComfyUI Manager (search ComfyUI-DrawThingsWrapper) or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/JosephThomasParker/ComfyUI-DrawThingsWrapper
cd ComfyUI-DrawThingsWrapper
pip install -r requirements.txt

then restart ComfyUI.

Common issues & troubleshooting

The LoRA has no effect. Almost always the name. lora must be the exact Draw Things filename - Copy configuration, paste, use it verbatim. A near-miss silently does nothing rather than erroring.

It's applied but the result is weird. Wrong architecture. A LoRA built for a different base than the model in your pipeline won't behave - match the LoRA to the base model.

Too strong / washed out. Turn weight down. Stacking several LoRAs compounds, so if you're chaining a few Add-Lora nodes, ease each one back rather than running them all at 1.0+.

Nothing renders. This node doesn't render - it only edits the pipeline. Confirm DrawThingsGenerateFromPipeline is downstream and the dict actually flows through this node on the way there.