1/12/26 YOU MAY NEED TO UPDATE THE COMFYUI-GGUF NODE PACK TO USE LTX-2!!!
These are two workflows that I've been using for my setup.
I have 12GB VRAM and 48GB system ram and I can run these easily.
The T2V is set for the 1280x720 and usually I get a 5s video in a little under 5 minutes. You can absolutely lessen that. I was making videos in 848x480 in about 2 minutes. So, it can FLY!
This does not use any fancy nodes (one node from Kijai KJNodes pack to load audio VAE and of course the GGUF node to load the GGUF model), no special optimization. It's just a standard workflow so you don't need anything like Sage, Flash Attention, that one thing that goes "PING!"... not needed.
I2V is set for a resolution of 640x640 but I have left a note in the spot where you can define your own resolution. I would stick in the 480-640 range (adjust for widescreen etc) the higher the res the better. You CAN absolutely do 1280x720 videos in I2V as well but they will take FOREVER. Talking like 3-5 minutes on the upscale PER ITERATION!! But, the results are much much better!
Links to the models used are right next to the models section, notes on what you need also there.
This is the native comfy workflow that has been altered to include the GGUF, separated VAE, clip connector, and a few other things. Should be just plug and play. Load in the workflow, download and set your models, test.
I have left a nice little prompt to use for T2V, I2V I'll include the prompt and provide the image used.
Drop a note if this helps anyone out there. I just want everyone to enjoy this new model because it is a lot of fun. It's not perfect but it is a meme factory for sure.