RuneXX

/

LTX-2.3-Workflows

like 716

All the workflows have a GGUF model loader at the left side that you can connect as main model loader instead of the Diffusion Loader.

You can also find an I2V & T2V workflow with the GGUF model loaders already connected here: https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main

If you cant find it, or dont know how to connect, I'll make a screenshot ;-)

thx i try

hi replaced the models with gguf but the out put is this

Unsloth GGUF should work straight out of the box, just swapping the model loader. Odd.
Will try my end.

What quant did you use? Anything at Q4 or more should give nice results
And did you use DEV or Distilled version?

If using DEV you also need to activate the lora right below the model loader, to make the DEV model run with few steps (as the workflow has)

i used another work flow and this is the result but its 8 steps and i cant find step option to edit

this the models

So you have the DEV model as GGUF.
Right under the model loader in the workflow, there is a Load Lora loader.
Connect your GGUF loader to this lora loader, and enable the lora loader (CTRL+B when its selected)

You can use the distilled-lora-384 that you already have, and/or youn can use the distilled lora from here https://huggingface.co/Comfy-Org/ltx-2.3/tree/main/split_files/loras
(the one in the link is smaller size, and also version 1.1 so usually works better)

After connecting GGUF loader to the lora loader (and enable the lora) all should be ok.

got it thx

Alternatively you can get the DISTILLED gguf model from Unsloth or Quantstack.
when using Distilled version, you dont need the distilled lora..

but its 8 steps and i cant find step option to edit

LTX workflows usually have hardcoded sigma with 8 steps (Manual sigma node)

In my workflows I usually "hide" under the manual signa node, a different way to connect the sigma.
If you connect that you can increase the steps (but its usually not needed, unless complex scenes etc)

thx again man