This workflow is based on the official workflows shared by LTXV, with some cool tweaks and visual organization.
Hey guys, I got early access to LTXV's new 13B parameter model through their Discord channel a few days ago and have been playing with it non stop, and now I'm happy to share a workflow I've created based on their official workflows.
I used their multiscale rendering method for upscaling which basically allows you to generate a very low res and quick result (768x512) and the upscale it up to FHD. For more technical info and questions I suggest to read the official post and documentation.
My suggestion is for you to bypass the Upscale group initially, then explore with prompts and seeds until you find a good initial i2v low res result, and once you're happy with it go ahead and upscale it. Just make sure you're using a 'fixed' seed value in your first generation.
I've bypassed the video extension by default, if you want to use it, simply enable the group.
To make things more convenient for me, I've combined some of their official workflows into one big workflows that includes: i2v, video extension and two video upscaling options - LTXV Upscaler and GAN upscaler. Note that GAN is super slow, but feel free to experiment with it.
Workflow .json file is attached to this post.
LTXV 13B Ultimate Workflow (Fast Results + Video Upscaling)
12

12