You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
LTX IC Lora Loader (Model Only) will not load into ComfyUi #553
I've deleted and re-cloned the Comfyui-LTXVideo custom node several times and it will just not link to the LTX IC Lora Loader (Model Only) node.
This obviously affects all workflows using IC lora's.
Any suggestions -please.
Thanks in advance.
HERE IS THE FIX THANKS TO GOOGLE GEMINI 3
This traceback confirms exactly why the loader won't launch.
The underlying computer vision library, Kornia, updated to version 0.8.x and removed an internal shortcut (pad) that the ComfyUI-LTXVideo extension relies on. Because this line of code crashes during startup, ComfyUI abandons the entire plugin, causing your node to appear as UNKNOWN.
You can resolve this error by applying a quick code fix or by downgrading Kornia via your terminal.Option 1: The Quick Code Fix (Recommended)This approach modifies the broken line directly so that it pulls pad from its modern location in PyTorch.
Go to this exact file folder on your computer:
E:\Ai\ComfyUi\ComfyUI\custom_nodes\ComfyUI-LTXVideo\
Open the file pyramid_blending.py using a basic text editor like Notepad.
Look near the top of the file (around lines 5–15) for this import statement:pythonfrom kornia.geometry.transform.pyramid import (
PyrUp,
build_laplacian_pyramid,
build_pyramid,
find_next_powerof_two,
is_powerof_two,
pad,
)
Use code with caution.
Delete the , pad line from that block, and add a new line immediately below it to import it directly from PyTorch:pythonfrom kornia.geometry.transform.pyramid import (
PyrUp,
build_laplacian_pyramid,
build_pyramid,
find_next_powerof_two,
is_powerof_two,
)
from torch.nn.functional import pad
Use code with caution.
Save the file and restart ComfyUI
Thanks, your remedy fixed it for me. I am just going to make a backup of this file into another location incase an update ruins again. Besides, in another year, this node probably deprecated and no longer needed or used.
Activity
Magenta-6 commentedon Aug 24, 2026
HERE IS THE FIX THANKS TO GOOGLE GEMINI 3
This traceback confirms exactly why the loader won't launch.
The underlying computer vision library, Kornia, updated to version 0.8.x and removed an internal shortcut (pad) that the ComfyUI-LTXVideo extension relies on. Because this line of code crashes during startup, ComfyUI abandons the entire plugin, causing your node to appear as UNKNOWN.
You can resolve this error by applying a quick code fix or by downgrading Kornia via your terminal.Option 1: The Quick Code Fix (Recommended)This approach modifies the broken line directly so that it pulls pad from its modern location in PyTorch.
Go to this exact file folder on your computer:
E:\Ai\ComfyUi\ComfyUI\custom_nodes\ComfyUI-LTXVideo\
Open the file pyramid_blending.py using a basic text editor like Notepad.
Look near the top of the file (around lines 5–15) for this import statement:pythonfrom kornia.geometry.transform.pyramid import (
PyrUp,
build_laplacian_pyramid,
build_pyramid,
find_next_powerof_two,
is_powerof_two,
pad,
)
Use code with caution.
Delete the , pad line from that block, and add a new line immediately below it to import it directly from PyTorch:pythonfrom kornia.geometry.transform.pyramid import (
PyrUp,
build_laplacian_pyramid,
build_pyramid,
find_next_powerof_two,
is_powerof_two,
)
from torch.nn.functional import pad
Use code with caution.
Save the file and restart ComfyUI
Magenta-6 commentedon Aug 24, 2026
THIS FIXED THE PROBLEM - BUT IF I TRY TO UPDATE ComfyUi-LTXVideo - The problem will come STRAIGHT ON BACK!
SJTAYLOR001 commentedon Aug 26, 2026
Nice, thank you.
Just had to do a clean Portable install after the usual fun of an update. Was having this problem.
UnrealGem commentedon Sep 1, 2026
Thanks, your remedy fixed it for me. I am just going to make a backup of this file into another location incase an update ruins again. Besides, in another year, this node probably deprecated and no longer needed or used.