Can you see where it Stitched? My Favorite LoRA of the year! Video Milestone + Efficiency Workflow

n this video, we break down SVI 2.0 Pro, the latest evolution of Stable Video Infinity, and explain why it fundamentally changes how infinite AI video generation works — especially when paired with Wan 2.2.

We will review the SVI 2.0 Pro model, that just got released right after Christmas. the SVI team + Kijai has been working closely to get us this new tool that is top tier in my opinion.
Furthermore, I've reviewed the whole structure of my own workflow and defined a bran new one to adapt to the need of SVI and how video are generated in parts, which allows to be as efficient as possible to create that long video:

Why is this better than before?
With SVI Pro, it's using latent to transition from clip to clip, and not last frame decode / encode, which aggravates the degradation. Using latent also include multiple motion data, which ensure smoothness of action transitions. Check out the video yourself to see if you can find where the "cut" happened.

Using latest Wan 2.2 model
Seamless Transition of Clips
Infinite Video potential
ID Character Consistency Maintained

💾 Download and test my ComfyUI workflow (FREE):
https://github.com/Jaevlon/AIArtistry...

☕ Support My Work:
If you find this tutorial helpful, consider buying me a coffee to keep me going and going...
https://buymeacoffee.com/jevlon

🚀 Tools Used:

ComfyUI:
Download & Installation
https://github.com/comfyanonymous/Com...

Wan Models Files (get the "Wan 2.2 I2V" files)
https://huggingface.co/Kijai/WanVideo...

Stable Video Infinity (SVI 2.0) models: (get the version 2.0 for Wan 2.2 high and low)
Select the PRO version
https://huggingface.co/vita-video-gen...

LoRAs:
Lightning LoRA: https://huggingface.co/Kijai/WanVideo...

Clip Encoder (umt5):
https://huggingface.co/Kijai/WanVideo...

WAN VAE:
https://huggingface.co/Kijai/WanVideo...