This is an offshoot of my work on the "not SORA" video generator. My ultimate goal is to create a universal tool that will let you use ADLCM+SVD+Zeroscope with pupeteering, create logical scenes then upscale to 4k native.

Currently it integrates:

1. Zeroscope v2 (T2V, V2V, upscale via model+SUPIR)

2. Modelscope v2 nodes (with SD15 diffuser)

3. ADLCM upscale

4. SDXL lightning upscale

If you make an amazing video with this, please post it on my discord at tinyurl.com/URSIUM - we'd love to see your work!

If you or your studio find this useful, consider a donation at https://ko-fi.com/stephantual to help me continue my work, as this is a huge project and will be released in phases. All future updates will be on this page.

The LORA for MS resolution normalization was compiled by exponentialML (the creator of the new SD15 + modelscope nodes). It's included as a zip file on this site.

Important:

• This is not a beginner project - but it's a great way to learn!

• The horror film look is just something I like. You can just switch from lcm>squarert_linear to lcm for 'normal' (real life) results.

• I don't always use comfyUI manager - if comfy complains about missing nodes, please download them on github from the main branches (links in the video description)

• This is bleeding edge (really) - things will change, break and have bugs. This was compiled on March 14th 2024. If you find a bug in any of the nodes, please report them to the indivual creators on their respective github to help the community.

• This was built and run on windows 11 with 24gb ram. I tested it to successfully run at 16gb if you drop the resolution to 720p and drop the FILMVFI clear cache to 5, SUPIR steps: 8. The quality will be lower, evidently.

Do not ask for support here. Please use the discord at tinyurl.com/URSIUM in the correct forum category. Please understand this is released 'AS IS' and comes with no support or warranty of any kind. Everything is released under CC-BY - meaning you can do whatever you want with it, but please give attribution to https://www.youtube.com/@stephantual