cmkflow/cmk_nodes

CMK Flow is a modular add-on package of custom nodes for ComfyUI, a visual tool people use to build AI image generation pipelines by connecting blocks together. The project follows two ideas at once: a closed, guided set of connected pipe style nodes meant to be used together in a specific way, and a separate collection of open nodes meant for free experimentation. The main way to use it is through the CMK Flow Browser, which shows the available modules, building blocks, and curated example workflows right inside ComfyUI. A typical pipeline flows from an image source, through an optional ControlNet step, an initial sampling pass, a refining pass, a detail enhancement step, a face processing step, and finally an upscale and save step. The project was inspired by another open source ComfyUI browser interface for managing models and images, though CMK Flow is a separate, unrelated project. Key features include a strict separation between model resources, process state, image data, and logs, plus custom connection interfaces designed to prevent nodes from being wired together incorrectly. It supports running multiple detail enhancement and face processing steps in parallel, and it caches unchanged parts of a pipeline so they do not need to be reprocessed every time. An image input node crops incoming images to keep their aspect ratio intact rather than stretching them, and a separate loader can load a target and source image side by side, only resizing the target. To install it, you clone the repository into your ComfyUI installation's custom_nodes folder and install its Python requirements, then restart ComfyUI. The tool requires ComfyUI's newer Vue Nodes interface to display its dynamic controls correctly, and depending on which features you use, it needs external AI models such as face detection, segmentation, and face restoration models, which are not included and must be downloaded separately. The face swapping features include a mandatory built in safety check that runs locally on your own computer. It scans source and target images before any face swap happens and blocks the operation if it detects explicit content or if either person appears to be underage, with no way to turn this check off in the public version. The developer is explicit that this is a technical safeguard, not a legal age verification system, and mistakes are possible.