Instructions to use Kijai/LTX2.3_comfy with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusion Single File
How to use Kijai/LTX2.3_comfy with Diffusion Single File:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
AI UGC - glitching with the eyes and face
Hey, I've been experimenting with LTX 2.3 to do some AI UGC video product reviews, and I can't seem to solve the issue of the eyes floating and looking "pasted-onto" the face.
I've tried realism loras, character loras, and even custom nodes and running it in python.
Also, it seems that when characters turn around, or even face 90 degrees, the entire video falls apart. I've been working on this for a while, anyone have opinions or suggestions?
Could perhaps try add more steps. But then you would need to use a basic scheduler type of node, where you can set steps (instead of the hardcoded manual sigmas).
Sometimes those few 8 steps with the distilled model can struggle with fast motion. 8 steps is extremely few steps
Other than that, if using distilled, make sure you use v1.1 version of distilled model or distilled lora. Seems to be a bit better
And higher resolution often does wonders. 1080p or 720p