TL;DR: We shipped a ComfyUI node pack that puts Civitai's tools right in your graph. About 160 nodes across image, video, audio, text, analysis, and training. Pick models, LoRAs, and ControlNets through loaders with built-in Civitai browsing, and pull your whole generation history back onto the canvas from a sidebar.
Why we're doing this
ComfyUI is where you build the real workflows. But wiring Civitai into one has always meant copy-pasting AIR strings, hunting model IDs, and gluing things together by hand.
So we made nodes. Real ones, typed and labeled, that drop into your graph and just work. The whole Civitai toolbox, in the place you already build.
How it works
Step 1. Clone the pack into your custom_nodes folder and install the one dependency:
cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txtStep 2. Sign in with your Civitai account from the sidebar, or set CIVITAI_API_TOKEN with an API key if you're running headless.
Step 3. Open the node menu, find the Civitai category, and start wiring. Outputs come back as native Comfy types (IMAGE, VIDEO, AUDIO), so they connect straight into the rest of your graph.
What you get
~160 nodes, one menu. Image, video, audio, text, analysis, and training, all under the Civitai category. Text-to-image, upscaling, background removal, video gen and interpolation, text-to-speech, transcription, captioning, tagging, moderation, resource training, and more.
Models as typed sockets. Pick models, LoRAs, embeddings, and ControlNets through loaders with built-in Civitai browsing. No pasting AIR strings into text widgets and hoping you got it right.
Native Comfy outputs. Everything returns as IMAGE, VIDEO, or AUDIO, so Civitai nodes mix freely with the rest of your graph.
Always current. The nodes are generated straight from our API spec, so they stay in sync as we add models and capabilities. No chasing manual updates.
Your history, in a sidebar. Every generation you've made on Civitai shows up in a panel. Filter by media type, paginate, and pull any result back in: add to canvas with loaders auto-wired, fill a selected loader, or drag the thumbnail on.
Where to find it
The pack lives on GitHub: github.com/civitai/civitai-comfy-nodes. Clone it, sign in, and the Civitai category shows up in your node menu. It's also available on the comfy registry.
Still cooking
These are live and ready to use now, but we're still tweaking and improving. Found a rough edge, or want something added? Open an issue or a PR on the repo. We're reading.
Thank you
You've been gluing Civitai into ComfyUI by hand for a long time. The least we can do is give you the nodes. Clone it. Wire it up. Build something.

