Lora Prompt Concatenation
Your full prompt with LoRA tags attached, ready for Civitai metadata
- lora_loader
- positive
You stacked three LoRAs in rgthree's Power Lora Loader, the image came out great, and then you uploaded it to Civitai and the generation info doesn't mention a single LoRA. That's not you failing - Power Lora Loader is a model node, and save nodes that write Civitai-compatible metadata never see the LoRA tags it's applying. Lora Prompt Concatenation closes exactly that gap: it takes your positive prompt, appends every enabled LoRA as a proper <lora:name:strength> tag, and hands you back one complete string that's ready for ComfyUI-Image-Saver (or any metadata field).
If you post to Civitai with the "show generation info" feature on, this is the node that makes the records accurate.
How it works
It's the pack's usual graph-introspection move, applied twice. The node parses the embedded workflow, traces its lora_loader input back to the Power Lora Loader, and reads that loader's lora_* widgets - keeping only the entries whose on toggle is enabled, and formatting each as <lora:name:strength>. Then it appends that run of tags directly onto the positive string you fed in. Disabled LoRAs are left out, and the strengths come straight from the loader's sliders, so the output always matches what actually ran.
It's strict about its input too: connect anything other than a Power Lora Loader (rgthree) to the loader socket and it throws Please check if Power Lora Loader (rgthree) is set as input. The node knows exactly what it's for.
The inputs that matter
Only three things touch this node:
- lora_loader (
*) - your Power Lora Loader. Theontoggles here decide which LoRAs appear in the output. - positive (STRING) - your positive prompt text, the one you'd normally wire into a CLIP Text Encode. This is a forced input, so it wants a real connection, not a typed box.
- positive (STRING) - the output, same name as the input, holding
your prompt <lora:name:strength><lora:name2:strength2>....
Wire that output into ComfyUI-Image-Saver's positive-prompt or metadata slot and the saved PNG carries the LoRA info Civitai wants. It also works fine for any text use - a display node, a filename template, a text file writer.
Install and things to watch
Part of GPS' Supplements for ComfyUI (Goshe-nite), which installs with no dependencies and no model files. ComfyUI Manager → search "GPS' Supplements for ComfyUI", or:
cd ComfyUI/custom_nodes
git clone https://github.com/Goshe-nite/comfyui-gps-supplements
Restart, then look under GPSupps.
Two notes. First, this node needs rgthree-comfy installed and a Power Lora Loader in the graph - there's no sensible standalone use, so install both packs together. Second, it reads the recorded workflow state, so if a value upstream is wired rather than typed, you get the wired result; if the graph's embedded JSON is missing (bypasses, API-only execution), you'll hit the pack's familiar Unable to get node info. For normal UI use, none of that bites - it just makes your uploads tell the truth.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| lora_loader | * | — | |
| positive | STRING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| positive | STRING | — |