Paste an error to find the matching fix, or explore workflows and the files behind them — all in one connected library.
Error logs open the Error Finder (nothing is stored or put in the URL). Anything else runs a search.
Releases, announcements and registry changes from the ComfyUI project.
The issues new ComfyUI users hit first, and how to get past them.
A first orientation to ComfyUI’s node canvas, workflow graph, model and node panels, execution queue, and saved outputs. Labels vary between versions; this guide uses the current official interface documentation and is not a live UI test.
Identify the format and source of JSON, images or videos before installing nodes and weights. Static inspection helps planning but cannot replace execution in the target environment.
Log keyword: Missing Node Types
Use this guide when an imported workflow contains unresolved nodes or the backend cannot find a node type. Save a workflow copy and its original source first. An unresolved node does not prove that all of its files are missing from the computer.
Log keyword: Value not in list
A saved ckpt_name, lora_name or another selection is absent from the current allowed list. Check the input name first: this validation error can also concern non-file options such as samplers.
Log keyword: CUDA out of memory
Identify the stage that requested GPU memory before reducing the workload. Advertised VRAM, weight-file size and the memory available to this particular task are different quantities.
Log keyword: ModuleNotFoundError
A successful installation in a terminal does not show that ComfyUI uses the same Python. Separate the import name, distribution name and running interpreter before changing dependencies.
Each route manages Python differently. Pick yours first, then follow the matching guide.
Use the official app on supported Windows, Apple silicon Macs or Linux systems; check the requirements for your platform.
Choose Desktop or a separate manual Python environment, and match PyTorch to the Mac.
Compare official Comfy Cloud with a self-hosted GPU, container or notebook.
A package with its own Python and ComfyUI folders. Learn which interpreter it uses before installing any dependency.
Plan the supported interpreter, device backend and a separate working directory instead of one pip command for every machine.
An independent reference for a Windows launcher: platform support and precautions before it changes your environment.
Which files each family needs, where they go, and a first workflow to run.
Core image techniques first, then memory, debugging and API integration.
Use the official basic image-to-image path to understand how input pixels influence a result. Denoise and seed serve different purposes; there is no single best value for every image.
A painted mask is only part of an inpainting workflow. Check the image, mask interpretation and supported model/encoding path together.
A larger image can mean new content beyond the frame or more pixels representing the existing composition. Choosing the wrong workflow may produce the requested dimensions but the wrong content.
Identify the expected edge, pose or depth representation and trace how its conditioning reaches the sampler. Different control models are not one universal node.
Start with one compatible LoRA and a controlled comparison. A populated selector does not establish that the sampling path uses the modified weights.
Downloading a repository is only one step. Identify the owning package, satisfy its dependencies and confirm that the running backend actually registers the required class.
Establish the runnable range of a particular setup, not a universal minimum-VRAM claim derived from a model name.
Replace random uninstalling with controlled groups and repeatable input. The goal is the smallest triggering set, not a blanket verdict on all third-party nodes.
This guide concerns an instance you own or are authorized to use. ComfyUI's native HTTP API is distinct from this site's knowledge MCP and does not authorize access to third-party services.
After execution, identify which nodes produced results and where those results reside. Node IDs, prompt IDs and filenames are different identifiers.
Use an explicit task and result contract instead of assuming an application entry script is a reusable library.
Check visible devices and the actual execution path before assuming that multiple GPUs pool memory.
Diagnose, inspect and connect without installing anything.
Pick a section by what you need to do right now.
Short answers, each with the full article behind it.
A node canvas holding the workflow graph, panels for models and nodes, an execution queue, and saved outputs. Labels vary between versions.
Read the full articleA successful install in a terminal does not show that ComfyUI runs the same Python. Check the import name, the package name and the running interpreter — Windows Portable has its own.
Read the full articleFirst find the stage that failed — model loading, sampling or VAE decoding — then reduce that load. Advertised VRAM, file size and the memory available to the task are different numbers.
Read the full articleCheck the scanned folder, the file name, which instance is running and any stale saved value. “Value not in list” can also concern non-file options such as samplers.
Read the full articleSave a copy first, then tell apart an absent package, a failed import and an unregistered class. A missing node does not prove the files are absent from your computer.
Read the full articlelocalhost means the device making the request. Confirm where ComfyUI actually runs — this machine, a container, a notebook or a remote server — before copying a URL.
Read the full articleNot first. It can mean a CPU-only PyTorch build, or a node assuming CUDA on a non-NVIDIA machine such as a Mac; the fixes differ.
Read the full articleYes, through the supported extra_model_paths configuration. It shares files only — not Python environments, installed nodes or workflow versions.
Read the full article
