CraterView.ai nodes for ComfyUI

Use Enhance, Denoise, Restore, Small face enhancement and Headshot as nodes in a ComfyUI workflow. Each node takes an image and returns an image. The work runs on our servers, so you need no graphics card and no model files.

Run Enhance on the original photograph, before Denoise or Restore. It might not work well on another model's result.

Install

In ComfyUI Manager, search for CraterView.ai, install it and restart ComfyUI. Or use comfy-cli:

comfy node install craterview

Or install it by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/craterviewai/craterview-comfyui
pip install -r craterview-comfyui/requirements.txt

Add your API key

Get a key from your dashboard. Making one costs nothing, and without credit or a subscription your jobs run free on the community queue. Then do one of these:

• Set a variable where ComfyUI starts, then restart it: export CV_API_KEY=cv_...
• Save it in a file, which needs no restart. Put the key in ComfyUI/user/__craterview/api_key.txt, creating the folder if it is not there. ComfyUI does not serve that folder over its web interface. If your ComfyUI keeps its folders somewhere else, run a node with no key: its error, under View details, names the exact path.

Never type the key into a node. ComfyUI saves every node input in each PNG it writes, so the key would go with every image you share. These nodes have no input for it, and saved images and workflows do not contain it.

In the cloud

• A rented GPU server: install with ComfyUI Manager or by hand, and set CV_API_KEY in the server's environment settings.
• Your own Docker image or serverless worker: git clone and pip install in the Dockerfile, and pass CV_API_KEY as a secret when it starts. Never build the key into the image.
• A hosted ComfyUI site that lets you add nodes: install with ComfyUI Manager, and use the site's environment settings, or the key file if it has a file browser.
• Comfy Cloud: not available yet. Comfy Cloud runs only the nodes Comfy has installed.

How a node behaves

• Each image in a batch is a job of its own. A batch of 16 is 16 jobs, each priced as one. Prices and limits are on each feature page.
• Without credit or a subscription, a job runs free on the community queue: the same models, with smaller size limits and longer waits at busy times.
• Running again with nothing changed starts no job. ComfyUI reuses the result. Changing anything before the node starts a new job, and so does restarting ComfyUI.
• A saved workflow keeps its settings when a model gains a new one, in ComfyUI 0.44 and later.
• Stop does not cancel a job. It stops ComfyUI waiting. The job carries on and is charged if it succeeds, and the log names it.
• A setting left at its starting value is not sent, so the model's own default applies. To set a result's size, change one of scale, width or height. Changing two is refused.
• To work on part of an image, crop it first with an image node.
• Images leave your machine. See the privacy policy for what is stored and for how long.

A walk-through with a first workflow is in the guide to the clients. The source and every detail are in the pack's repository.