Downloads · 30 days
37
10% of all-time downloads
kabachuha/ltx23-paste
ltx23-paste is a image-to-video model from kabachuha. Use it for the image-to-video task on the model card, and read the license before you ship it in a product. It is set up for diffusers. The card lists the license as other.
Downloads · 30 days
37
10% of all-time downloads
All-time downloads
389
Public
Repo size
696 MB
Likes
5
Public
Click a slice to open those files.
.safetensors654 MB · 94%
From the Hugging Face model README
Squeeze your characters from a tube like liquid paste!
BIG NOTE: This is a Last frame to video LoRA and you need to use frame "guides". They are in native Comfy, so it won't add load.
The workflow is in the videos metadata or inside kabapaste-workflow.json; at my Huggingface page.
As you may notice, the videos are at 50 FPS. Otherwise, the workflow is fully default, but with 30 steps and distill lora strength at 0.6 for reportedly higher quality. It uses KJ nodes for image resize and VHS for metadata saving.
For training details, see my other recent 2.3 LoRAs on Huggingface.
When uploading the examples, a big bruh moment occured in hindsight - I uploaded the videos where the generating model was actually 2.0, despite the LoRA targeting 2.3 in its training. It may be confusing, so I updated the json workflow to 2.3 and left the old one as 2.0. The 2.3 videos generate just fine or even better, I attached them into the folder 'video_examples-ltx23'. I left the gallery above untouched for Civitai expectations compatibility.
Trigger phrase is:
KABAPASTE The girl is squeezed from a tube like paste.
Download it in the Files & versions tab.
(See bruh notice)
<video width="480" height="480" controls> <source src="https://huggingface.co/kabachuha/ltx23-paste/resolve/main/video_examples-ltx23/3-street.mp4" type="video/mp4"> </video>