CivitAI SDK API Nodes for ComfyUI

Custom nodes for ComfyUI that generate images through the CivitAI Orchestration API.

Generate images in the cloud using CivitAI checkpoints and LoRAs by AIR URN — no local model files required.

Nodes

Node	Description
CivitAI Generator	Sends a text-to-image request with API key, checkpoint AIR, and generation parameters. Returns a ComfyUI IMAGE.
CivitAI LoRA	Adds a LoRA by AIR URN with configurable strength. Supports chaining multiple LoRAs.

Workflow example

[CivitAI LoRA] ──► [CivitAI LoRA] ──► lora_stack ──► [CivitAI Generator] ──► IMAGE

1. Add one or more CivitAI LoRA nodes and chain them together.
2. Connect the final lora_stack output to CivitAI Generator (optional).
3. Set your API key, checkpoint AIR, prompt, and other parameters.
4. Connect the image output to Preview Image or Save Image.

Installation

Via ComfyUI Manager

Search for CivitAI SDK API nodes for ComfyUI (if listed) or install from Git URL.

Manual install

cd ComfyUI/custom_nodes
git clone https://github.com/Ave-Mollan/CivitAI-SDK-API-nodes-for-ComfyUI.git
pip install -r CivitAI-SDK-API-nodes-for-ComfyUI/requirements.txt

Restart ComfyUI after installation.

Requirements

• ComfyUI
• Python 3.10+
• Dependencies from requirements.txt:
  • requests
  • Pillow

torch and numpy are provided by ComfyUI.

API key

1. Open CivitAI Account Settings.
2. Create an API token.
3. Paste it into the api_key field of the CivitAI Generator node.

Note: Generation uses CivitAI Buzz credits. Make sure your account has sufficient balance.

AIR URN format

Models on CivitAI are referenced by AIR (Asset Identifier Resource) URNs:

urn:air:sd1:checkpoint:civitai:4384@128713
urn:air:sdxl:checkpoint:civitai:827184@2883731
urn:air:sdxl:lora:civitai:1105685@1242203

Find the AIR URN on the model page on CivitAI (AIR section) or in the model documentation.

CivitAI Generator parameters

Parameter	Description
api_key	CivitAI API token
model_air	Checkpoint AIR URN
prompt	Positive prompt
negative_prompt	Negative prompt
width / height	Image dimensions (256–2048, step 64)
steps	Sampling steps (1–150)
cfg_scale	CFG scale (1.0–30.0)
scheduler	Sampler scheduler (EulerA, DPM2, DDIM, etc.)
seed	Seed (-1 = random)
clip_skip	CLIP skip layers (0 = disabled)
timeout	Max wait time for generation in seconds
lora_stack	Optional chained LoRA stack

Outputs: image (ComfyUI IMAGE), job_info (token, job ID, image URLs)

CivitAI LoRA parameters

Parameter	Description
lora_air	LoRA AIR URN
strength	LoRA strength (-2.0 to 2.0)
lora_stack	Optional input to chain another LoRA

Supported models

The extension auto-detects the base model from the AIR URN:

• SD_1_5 — SD1.x checkpoints and LoRAs
• SDXL — SDXL, SD3, Flux checkpoints and LoRAs

Troubleshooting

Issue	Solution
API key is required	Enter a valid CivitAI API token
Invalid AIR format	Use full URN starting with urn:air:
Generation timed out	Increase timeout or check CivitAI service status
Job finished without image output	Check Buzz balance, model AIR, and prompt for policy violations

License

MIT License — see LICENSE.

Links

• CivitAI Developer Docs
• CivitAI Python SDK
• ComfyUI