Breaking Changes: 22th February 2026: Model loading change
See more here: https://huggingface.co/Kijai/LTXV2_comfy
Will update the workflow asap to reflect the breaking changes in ComfyUI model loader logic.
And if you have some experience with nodes, see the link above on what to change, if you want to do yourself. (basically just swapping out the main model loader with a new one)
The workflows are based on the extracted models from https://huggingface.co/Kijai/LTXV2_comfy The extracted models runs easier on the computer (as separate files), as well as GGUF support etc
(but you can easily swap out the model loader for the ComfyUI default model loader if you want to load the checkpoint with "all in one" vae built-in etc)
Gemma 3 12B it GGUF text encoder: https://huggingface.co/unsloth/gemma-3-12b-it-GGUF/
Needed nodes:
https://github.com/kijai/ComfyUI-KJNodes (NB! Must be up to date for LTX-2 support)
https://github.com/city96/ComfyUI-GGUF (NB! Must be up to date for LTX-2 support)
ComfyUI itself must be updated to very latest
(video made with LTX-2, Credit to https://www.reddit.com/user/fantazart/) https://www.reddit.com/r/StableDiffusion/comments/1qeovkh/ltx2_cinematic_love_letter_to_opensource_community/
A general guide: https://docs.ltx.video/open-source-model/integration-tools/comfy-ui
More workflows:
ComfyUI official workflows: https://docs.comfy.org/tutorials/video/ltx/ltx-2
LTX-Video official workflows: https://github.com/Lightricks/ComfyUI-LTXVideo/tree/master/example_workflows
Some really nice clean workflows here: https://comfyui.nomadoor.net/en/basic-workflows/ltx-2/
RunComfy (can download workflow to use locally):
LTX-2 Controlnet (pose, depth etc) https://www.runcomfy.com/comfyui-workflows/ltx-2-controlnet-in-comfyui-depth-controlled-video-workflow
LTX-2 First Last Frame https://www.runcomfy.com/comfyui-workflows/ltx-2-first-last-frame-in-comfyui-audio-visual-motion-control