LogoWonderful Launcher
  • Home
  • Pricing
  • Docs
  • Download

ComfyUI Plugin Import Failed: Fix Custom Node Errors

VerifiedMedium riskTested on Windows 10, Windows 11 | Launcher 1.x | ComfyUI portable

Fix ComfyUI plugin import failed errors, broken custom node imports, red workflow nodes, and dependency drift without reinstalling your environment.

Tested Environment

  • OS: Windows 10 / 11
  • Launcher: Wonderful Launcher v1.x
  • ComfyUI: Portable / Managed install
  • Python: 3.11+
  • CUDA / Torch: CUDA 12.x / Torch 2.x
  • Last tested: 2026-05-19

If the startup log shows IMPORT FAILED, a custom node was downloaded but the Python process that starts ComfyUI could not import it. Identify the exact plugin and traceback first, fix the earliest import failure (not the red node you noticed later), and install packages only in the Python environment that launches ComfyUI. This is a dependency or packaging problem first — not proof that all of ComfyUI is broken.

If the visible error is different, use the matching guide instead: Failed to Save Workflow Draft → save the workflow JSON first; failed to fetch server logs → Failed to Fetch Server Logs; Manager failed to get custom node list → Failed to Get Custom Node List.

Typical log lines:

IMPORT FAILED: ComfyUI-ExampleNode
ModuleNotFoundError: No module named 'somepackage'
IMPORT FAILED: ComfyUI-AnotherNode
ImportError: DLL load failed while importing cv2
IMPORT FAILED: ComfyUI-SomePlugin
AttributeError: module 'torch' has no attribute '...'

Separate three symptoms that get mixed together

SymptomWhat it usually means
Red node in a workflowThe workflow asks for a node class that is not registered
IMPORT FAILED during startupA plugin folder exists, but Python could not import it
Error only when queueing a promptThe node exists, but runtime inputs, models, CUDA, or a backend failed

Fix the earliest failing layer first. Installing packages for a runtime error will not help if the plugin never registered its nodes.

Triage: package-specific routes

If the traceback names one of these, use the narrow guide before running broad plugin requirements:

Log clueUsually affectsSafer next guide
No module named 'insightface'ReActor, InstantID, IPAdapter FaceIDInsightFace Missing
No module named 'onnx' or onnxruntimeDWPose, ReActor, ONNX inferenceONNX / ONNXRuntime Missing
No module named 'gguf'ComfyUI-GGUF loader nodesComfyUI-GGUF
No module named 'triton'SageAttention, compiled kernels, some video workflowsTriton Missing or Unavailable
No module named 'sageattention'Optional acceleration or video workflowsSageAttention Missing
No module named 'llama_cpp'QwenVL GGUF, local LLM/VLM, prompt enhancersllama_cpp missing
No module named 'nunchaku'Nunchaku FLUX, Qwen-Image, SVDQuantNunchaku Missing
Torch not compiled with CUDA enabledGPU PyTorch build replaced or inactiveTorch CUDA repair
AttributeError against ComfyUI internalsPlugin source no longer matches your ComfyUI versionUpdate plugin or patch source, not packages
Easy Use, Layer Style, rgthree, VideoHelperSuite import failedPopular pack installed but not loadingPopular custom node packs

Step 1: Confirm which plugin actually failed

Do not start by installing random packages. Answer: which plugin failed, and what error type stopped the import? Check in order:

  • Startup log: find IMPORT FAILED and copy the full traceback.
  • Import failure API: if the UI opens, visit http://127.0.0.1:8188/v2/customnode/import_fail_info_bulk for a JSON list of failures.
  • Workflow symptoms: note whether one plugin family is missing, multiple unrelated plugins are missing, or a plugin worked before another install changed the environment.

Step 2: Classify the failure before you fix it

  • Missing package (ModuleNotFoundError: No module named 'onnxruntime') — the plugin expects a package that is not installed.
  • Broken binary package (ImportError: DLL load failed while importing cv2) — the wheel does not match your Python, CUDA, or Windows runtime.
  • Core dependency drift (AttributeError: module 'torch' has no attribute '...') — another plugin install changed torch, numpy, pillow, or opencv-python.
  • Repo or source problem — the repo moved, cloned partially, left Git LFS pointers, or is too old for your ComfyUI version.

Before any install command, check: did the plugin ever work here; is the folder present but broken or completely missing; did another plugin install happen right before this started failing.

Step 3: Repair with the narrowest fix

If a package is missing, install it in the Python that starts ComfyUI:

Install typeSafer command
Official GitHub Windows portable packageFrom the portable package root: .\python_embeded\python.exe -s -m pip install somepackage
Manual Git + venv installActivate the venv, then run python -m pip install somepackage
ComfyUI Desktop or managed launcherUse the app's environment/terminal tools. Do not assume the portable python_embeded layout exists.

If you suspect dependency drift, run pip check and look for conflicts around torch, torchvision, torchaudio, numpy, pillow, opencv-python, transformers, diffusers. Repair the core package first, then retest. See ComfyUI Dependency Conflicts.

If the plugin itself is outdated or partially broken, verify the repo is correct, the source files are real code (not Git LFS pointers), and the plugin still supports your ComfyUI version. If needed, cd custom_nodes/<plugin-name> and git pull, then reinstall only that plugin's requirements:

Do not run pip install -r requirements.txt blindly.

  1. Confirm which Python environment is active (where python or which python)
  2. Inspect requirements.txt — check if it pins torch, numpy, or opencv versions
  3. Install only the missing package when possible: pip install <package-name>
  4. Run pip check after installation to verify no conflicts were introduced

For the official Windows portable package, run from the portable root with .\python_embeded\python.exe -s -m pip install -r .\ComfyUI\custom_nodes\<plugin-name>\requirements.txt.

Verification ladder

A plugin folder, a successful clone, or an offline Python import is not the same as a working ComfyUI node. Confirm each rung before calling the repair done:

Verified stateWhat it provesWhat it does not prove yet
Plugin folder existsFiles are present under custom_nodes/The plugin imported successfully
python -c "import package" worksOne package imports offlineComfyUI used the same Python or registered the node
Startup log has no IMPORT FAILED for the pluginThe import got past startupThe workflow has all model files and runtime inputs
/v2/customnode/import_fail_info_bulk is clean for that pluginNo recorded import failureThe workflow will execute without model/CUDA/input errors
Red nodes are gone after restartNode classes registeredModel files and runtime backends may still be missing

If nodes now register but the workflow asks for model files, switch to model placement or cannot find model instead of reinstalling the plugin.

When this is really a dependency-conflict problem

If one plugin import failure turns into several unrelated plugin failures after a package install, switch to:

  • ComfyUI Dependency Conflicts
  • Repair ComfyUI Portable Dependencies One Missing Module At A Time

Related guides

  • ComfyUI Startup Failed? How to Diagnose and Recover Faster
  • ComfyUI Dependency Conflicts
  • ComfyUI No Module Named Error: When It Is Safe to Ignore
  • ComfyUI No Module Named 'llama_cpp'
  • ComfyUI Plugin Management
  • Popular ComfyUI Custom Node Packs
  • Troubleshooting Decision Tree

Source References

  • ComfyUI troubleshooting overview
  • ComfyUI custom node troubleshooting guide
  • ComfyUI model troubleshooting guide
  • ComfyUI Manager installation guide
Related guides:Installed custom nodes and broke ComfyUI?ComfyUI startup failedComfyUI dependency conflicts

Start free with Wonderful Launcher if this affects your real ComfyUI environment. It keeps launcher-native repair, task logs, and runtime checks in one place; credits are only for image generation and metered tools.

Download Wonderful LauncherSee credit packages

Did this fix your issue?

Your answer helps prioritize verified ComfyUI repairs.

Table of Contents

Separate three symptoms that get mixed together
Triage: package-specific routes
Step 1: Confirm which plugin actually failed
Step 2: Classify the failure before you fix it
Step 3: Repair with the narrowest fix
Verification ladder
When this is really a dependency-conflict problem
Related guides
Source References