A GGUF version of the Image to Video workflow of the newest LTX Video version 2.3 in subversion 1.1.
Howto
Drag an image into the image nodes, adjust the prompt, and press queue. The rest should fit.
Description
An image to video ComfyUI workflow with LTX Video 2.3 subversion 1.1. The nodes are wired and visible in the traditional way and grouped together. No set and get nodes, no subgraph. I like it simple. It produces two videos, one with audio, one without.
Time
The example video in a resolution of 1024 x 576 was generated in 240 seconds at Windows 11 at an AMD Radeon AI PRO R9700. I have yet to test it on Linux, which is usually faster by a third.
Requirements
This workflow was generated with 32 gb vram. There is a ram saver node involved, which you can turn on. So you should be able to go lower. The bottleneck that even made my rig freeze is the VAE decode.
Description
FAQ
Comments (6)
It really generates fantastic, true HD quality. I don't know what settings you mastered, but in fact it even improves the quality and sharpness of the source image when turning it to video. I am impressed! It is however about 4 minutes slower than my other workflows, that render 20 seconds for 10 minutes. But it's worth it.
Thanks. That's most probably the model. First it is the new subversion 1.1, and it adds quite a bit quality to use the Q5 gguf model over the Q4 one. But it is a bit slower then of course. For me the Q8 gguf model was not worth the tradeoff between quality and increased generation time. Q5 defintelty was.
Yeah, it’s a shame that the LTX 2.3 1.1 model, even compressed, is still way too heavy for low VRAM. Minimum 14 GB in Q3_K_S is ridiculous, especially when you see the crappy audio it outputs compared to Wan with audio input
@huguesbon590 - i think you look at wrong places, choose from here: https://huggingface.co/unsloth/LTX-2.3-GGUF/tree/main/distilled-1.1
I am at Q4_0 at the moment with 12 GB VRAM. It is way more stable than wan
Don't work for me:
"TypeError: AudioVAE.__init__() takes 2 positional arguments but 3 were given"
About
ComfyUI 0.21.1ComfyUI_frontend v1.45.7Templates v0.9.75DiscordComfyOrgEasyUse v1.3.7LoRA Manager v1.0.5-cc147a1ComfyUI-OllamaComfyUI-Manager V3.39.2rgthree-comfy v1.0.2512112053
:(
I would check the models first, one might be faulty. Or you have a not compatible version of LTX video node. For me it works fine. So i cannot say what the problem is, sorry.