KJ Nodes Dead in the Water #15204
Replies: 2 comments
|
Python Version PyTorch Version |
|
This looks like a KJNodes import failure, not evidence of a Python 3.12 clash. Start ComfyUI from the console and find the first KJNodes traceback during startup. That exception is the useful part; the red missing-node boxes in the workflow are only the aftermath. For Portable, install KJNodes requirements with Portable's embedded Python, not the system Python: python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-KJNodes\requirements.txtThen update ComfyUI core and KJNodes together and restart. KJNodes' current LTX module imports APIs from current ComfyUI core, so mixing an old core with a newer custom-node checkout can make the whole LTX block disappear. After restart, or try using my comfyui skill; https://github.com/vavo/comfyui-codex |
Uh oh!
There was an error while loading. Please reload this page.
I am a ComfyUI Portable user for years now on the following system since Dec 2025:
:: HARDWARE INFO ::
System RAM: 63 GB
GPU Model: NVIDIA GeForce RTX 5090
Video VRAM: 32 GB
:: SOFTWARE & VERSIONS ::
ComfyUI : v0.29.0
Windows long path support is Enabled
Python : 3.12
PyTorch : 2.10
CUDA Core : 13.0
NVIDIA-drv: 610.47
The issue I've been confronting is specifically with ALL LTX2.3 workflows using the KJ Nodes for the VAE Loader especially. Manager install, CMD install, update, version change, etc., nothing seems to "find" those components and even the ComfyUI Templates won't work no less others' workflows. I regularly update CUI and the nodes, also. I can run Wan2.2 workflows without a hitch all day but can't get LTX2.3 to cooperate.
Any ideas? Could it be a Python version clash or something similar?
All reactions