Fix Red Nodes in a ComfyUI Workflow
Academy is temporarily unavailable while it is being rebuilt.
Red nodes in ComfyUI usually mean the workflow references a node type your installation cannot find. This often happens when you download a community workflow that uses custom node packs such as ComfyUI-Impact-Pack, comfyui_controlnet_aux, ComfyUI_IPAdapter_plus, or ComfyUI-AnimateDiff-Evolved. The goal is not to rebuild the workflow blindly; it is to identify the missing node, install the matching pack, and test again.
Start by reading the red node label or the error message. Look for wording such as `Node type not found` and copy the exact node name. Exact names matter because different packs can contain similar-looking tools. For example, a missing `FaceDetailer` node points you toward ComfyUI-Impact-Pack, while `OpenPosePreprocessor` commonly points to comfyui_controlnet_aux.
If you use ComfyUI Manager, search for the missing node or the likely pack name, install it, then restart ComfyUI. Restarting matters because ComfyUI needs to reload Python modules and register the new node classes. If you install files manually instead, place the custom node folder in the correct custom nodes directory for your ComfyUI setup, then restart in the same way.
After restarting, reload the workflow and check the canvas before queuing a generation. A successful first check is simple: the former red nodes should display normally, and their input and output sockets should be visible. If a node is still red, either the wrong pack was installed, the installation failed, or the workflow depends on an additional custom node pack.
Once the red nodes are resolved, address connection errors one at a time. Links must join compatible data types: model to model, conditioning to conditioning, image to image, and so on. If validation reports a missing required input, inspect the specific field on the node. Some inputs need a connection; others are widget values such as text, numbers, or selected options. Set or connect the required input before changing generation settings.
The observable sign of success is that the workflow validates and reaches an output node such as Preview Image or Save Image. If it still fails, mute optional sections and test the basic text-to-image path first, then unmute one group of nodes at a time.