This is a combination of tools to automate a process I call txt2img2vid for ComfyUI.

I find you get much better control over your video if you can step by step go from txt2img, and then use that for img2vid through LTX.

Everything generated here is on a RTX4070 using about 10gb of vram, processing takes about 5 minutes total, but because I have to load and unload models from my tiny card, the whole thing takes about 10 minutes to make 10 seconds of video.

There is also an option to use your existing images for video gen in this workflow.

Install:

I installed Comfy using this guide here:

https://www.patreon.com/posts/1-click-install-121355091?utm_medium=clipboard_copy&utm_source=copyLink&utm_campaign=postshare_fan&utm_content=web_share

This lora is required to get the same style as my images:

https://civarchive.com/models/128417/age-slider

I recommend this SD1.5 model:

https://civarchive.com/models/209288/katafract

If you don't have the VAE, you can get it here:

https://civarchive.com/models/276082/vae-ft-mse-840000-ema-pruned-or-840000-or-840k-sd15-vae

Credits:

This uses pieces of this workflow for video gen:

https://civarchive.com/models/995093/ltx-image-to-video-with-stg-caption-and-clip-extend-workflow

And parts of this for image gen:

https://civarchive.com/models/59806/sd15-template-workflows-for-comfyui