LTX-2 GGUF I2V image-to-video workflow with power lora loaders and tiled vae decode, I modified this workflow to produce better quality output and used KJNodes to load the VAE and GGUF. detailer lora can be turned off or on in stage 2 depending if you want itLoad your camera loras on the subgraph and any sfw/nsfw loras you wish to use.
You can also try changing stage 1 scheduler from 20 to 40 steps to see how it changes video output quality
Description
first version of this workflow
FAQ
Comments (11)
Thank you!
Almost 3 weeks in and no T2V loras anywhere. We are doomed.
We need wan 2.5/2.6 now.
They will come, the training process takes time, wan 2.6 would be nice but I suspect its too large to run on consumer cards and I doubt alibaba will release it
@sexgod1979 Genau.
I recommend learning to meditate so that you don't get so worked up over such things.
Do you have any idea what causes the weird flicker at various points in the video? I'm seeing it at about the halfway point of one I'm generating, and I see it in your turkey and dinosaur examples, but don't notice it in the alien one
I think its the distill lora, im looking at that now, may also be the dolly camera lora as the alien one only uses the static camera. I haven't been happy with the camera loras. Do this test, only load the static camera lora, and do some type of static video, if theres is no flicker but you get a flicker when changing to dolly left/right than perhaps thats the issue. Trying to figure that out now
@sexgod1979 Seems to happen regardless, with just the static lora on or even all loras turned off
my image immediately turns to noise and the audio turns to static. have all models and loras loaded, happens with fresh workflow. no idea on this one :(
Try to update the ComfyUI with the ComfyUI-manager update all (not with the update.bat in the update-directory from ComfyUI !!!!). I had the same problem 😁
For me it runs perfectly. I have a RTX 4080 with 12GByte VRAM and 32 GByte "normal" RAM. Thank you for sharing it wiht us.