📥👑 Load Lora Pony (+Download from CivitAi)

The Pony-specific LoRA loader in Bjornulf's CivitAI family. You pick a Pony LoRA from a curated in-node gallery (around 77 entries - styles, backgrounds, control LoRAs), and the node downloads it from CivitAI if you don't have it, then applies it to your model and CLIP. It's a Load LoRA node with a CivitAI dropdown, and it also hands back the trigger words.

Why a Pony-specific one instead of just using the SDXL LoRA node? Because CivitAI treats Pony as its own base-model category. Pony diverged far enough from SDXL that SDXL LoRAs technically load but produce degraded results, and vice versa. Keeping Pony LoRAs in their own list (and their own folder) keeps you from mixing incompatible resources.

How it works

Pass in your model and clip, choose a LoRA, and on run the node checks locally; if the file's missing it downloads it from CivitAI into ComfyUI/models/loras/Bjornulf_civitAI/pony/, then patches the LoRA into your model and CLIP at your chosen strengths. Ordinary LoRA application - CivitAI is just how the file arrives.

The inputs and outputs that matter

• image - the LoRA picker, a thumbnail gallery of about 77 Pony LoRAs. Your main choice.
• model and clip - from your Pony checkpoint loader.
• strength_model and strength_clip - LoRA influence (default 1.0; the −20 to 20 range is wide, but 0.5–1.2 is the real world). Drop strength_model if it's overwhelming the gen.
• civitai_token - CivitAI API token, needed only for the download.

Outputs: model and clip to chain into your sampler, plus name, civitai_url, and trigger_words strings.

Getting it to actually work with Pony

A couple of Pony realities the node can't handle for you:

• You still need the score-tag preamble on the checkpoint prompt: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up. The LoRA doesn't change that.
• Feed the trigger word in. Wire the trigger_words output into your prompt - many Pony style/character LoRAs do nothing without their activation keyword.
• Keep it on a Pony base. A Pony LoRA on a plain SDXL checkpoint will misbehave; pair this with a Pony (or Pony-merge) checkpoint like AutismMix.

How to install it

ComfyUI Manager: search Bjornulf_custom_nodes, install, restart. Manually:

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

Restart ComfyUI. Downloads depend on civitai-py from that requirements file. Windows portable: .\python.exe -m pip install ....

Common issues

The download needs a valid token, and gated LoRAs can refuse - check the token first. A lot of Pony LoRAs on CivitAI are NSFW, and the list reflects that. If a LoRA seems inert, it's almost always a missing trigger word or a base-model mismatch. If it's overcooking, lower strength_model. Use the SDXL, SD1.5, or Hunyuan sibling nodes for those bases - this one is Pony only.