Resources you need:
📂Files :
For base version :
2B Model (7Gb) : ltxv-2b-0.9.6-dev-04-25.safetensors
In models/checkpoints
or
13B Model (29Gb) : ltxv-13b-0.9.7-dev.safetensors
In models/checkpoints
For GGUF version :
Quant Model : ltxv-13b-0.9.7-dev-QX_0.gguf
In models/unet
VAE : ltxv-13b-0.9.7-vae-BF16.safetensors
In models/vae
📦Custom Nodes :
Description
Base version
FAQ
Comments (23)
EDIT: Nevermind, updating comfyui and python dependencies instead fixed my issue.
updated comfyui and tried this. got error: SamplerCustomAdvanced
Only multiplication of row-major and column-major matrices is supported by cuBLASLt
I keep getting just tiny squares flashing on my results, tried playing around with various resolutions and settings. not sure why. i'll see if other workflows give me same issue and see if i can figure out why it's happening.
I have try it on Linux with L40S for the 14B and on windows with 4080 for 2B (using this version of comfyui https://civitai.com/models/1309415)
@UmeAiRTÂ I did get basic image to video quantized workfflow working from the comfyui-ltxvideo github. i had to install q8-kernals. so github workflow is working for me. but then tried going back to yours, using exact same model and i still get little squares as a result. I'm using the 13b model linked from your page.
Same problem.
@Testhamster Do you have the latest version of comfyui?
@UmeAiRTÂ Yep, hit the "Update ComfyUI" + "Update All" via the Manager :/
@Testhamster I managed to reproduce the bug, it only appears with version fp8 13B not base 13B, I'm working on it. (Try 2B in the meantime)
@Testhamster @zengrath I uploaded a new version with GGUF for 13B
@UmeAiRTÂ GGUF-Version works fine, thx mate!
@UmeAiRTÂ Can you also do it for the TXT2IMG version?
@genural I had already sent it, it was pending and seems available now.
tried both new versions of workflow now, gguf works for me. the base one still gives me a tiled scrambled result. will stick to gguf for now.
Hi there, I'm getting this error:
Unrecognized configuration class <class 'transformers_modules.Florence-2-base-PromptGen-v2.0.configuration_florence2.Florence2LanguageConfig'>.
I tried downgrading my Transformers version as suggested in several threads, but that doesn't seem to solve the issue. Any ideas, please?
if anybody is having any issues just ask chatgpt thats how i solve all my issues, just copy the report into chatgpt, and boom let it fly
I wish the description would add some details about what these workflows actually do. i.e. what is Florence?
This is just a workflow that allows you to create a video from an image. Florence is a model that allows you to do image to text to have an "automatic" prompt.
@UmeAiRTÂ Thanks! I guess I just don't know what LTXV is. I went through your I2V guide for your other workflow though, and I'm so thankful for how well explained it is and how clean that workflow is!
Your workflow doesn't recognize the GGUF CLIP, I've downloaded the exact clip, but for some reason, comfy isn't showing it in the list of clips.
Because this version use normal clip, i started adding the gguf clip loader to certain workflows but I'm traveling for work and I haven't had time for that one.
Got the Safetensor workflow working, but it gives me flashing squares.
This workflow is great but main comment is that the preview video has much better movement (more dramatic bouncing/jiggling/etc) and better, er, fluids than the output. Any ideas how I can improve movement etc. in the final output to be more like the preview?