Easy Apply LoraPrompt

If you came from Automatic1111 or Forge, you're used to just writing <lora:add_detail:0.8> in the prompt and having it work. ComfyUI normally makes you wire up a separate LoRA loader node for each one, which is more explicit but also more clicking. easy loraPromptApply brings the A1111 habit back: it reads the <lora:name:strength> tags right out of your prompt text, loads and applies those LoRAs to your model and CLIP, and strips the tags out so they don't pollute the actual conditioning.

It's a genuinely nice convenience for anyone who thinks in prompts rather than in nodes, and it plays well with wildcard-driven workflows where the LoRA you want might not be known until the prompt is populated.

How it works

You hand it a model, a clip, and your prompt string. It scans the string for inline LoRA syntax, resolves each tag to a file in your models/loras folder, and patches the model and CLIP with each LoRA at the strength you wrote. Then it returns the cleaned prompt with the tags removed - so downstream you encode a prompt that's just words, while the model has already had the LoRAs baked in. One node in, model/clip/prompt out, LoRAs handled.

The inputs and outputs that matter

• model and clip - the pair you want the LoRAs applied to. Come from your checkpoint loader.
• positive - the prompt text containing your <lora:...> tags. This is where the syntax gets parsed.
• negative (optional) - a negative prompt string; it's passed through cleaned as well.

Outputs are the four you'd expect: model and clip (now patched with the LoRAs - wire these into your sampler / CLIP encode), and positive / negative (the tag-stripped strings, ready to encode). The key mental model: this node's model and CLIP outputs are the ones carrying the LoRAs, so route those forward, not the originals.

How to install it

Part of ComfyUI-Easy-Use. ComfyUI Manager: search ComfyUI-Easy-Use, install, restart. By hand:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Use

then install.bat (Windows) or pip install -r ComfyUI-Easy-Use/requirements.txt, and restart. It uses the LoRAs already in your models/loras folder - nothing extra to download.

Common issues & troubleshooting

The LoRA silently doesn't load. The tag name has to match a real file. <lora:MyStyle:0.8> needs MyStyle.safetensors to exist in your loras folder - no extension in the tag, exact name match. A typo just means the tag gets ignored (or errors), and you'll wonder why the effect never showed up. The pack's own changelog notes a history of "LoRA not filled in completely" causing silent misses, so spell it exactly.

You wired the wrong model forward. The most common mistake: applying the node but then feeding your original checkpoint's model/clip into the sampler instead of this node's outputs. If the LoRA seems to do nothing, check that the sampler is downstream of this node's model and clip.

Tags leaking into conditioning. If you're not using this node (or a wildcard node that also parses LoRAs) and you left <lora:...> in a prompt going straight to a CLIP encode, those characters get treated as literal text and quietly muddy your generation. This node exists precisely to prevent that - let it strip the tags.

Strength written wrong. <lora:name:0.8> sets model strength; leaving the number off applies the default. If a LoRA is overpowering the image, lower that number rather than reaching for a different node.