Find me at JJANG
Trained on various stills from over a dozen old family VHS clips I digitized. Works best with realistic scenarios. Seems to do poorly with low light/night time images.
Trigger word is "vh5" Can increase strength slightly for heavier distortion.
I didn't bother removing the time stamp present in most of the images, so sometimes it can come through. If you write something like "text at the bottom left says 2:24 PM" it'll show perfectly.
Please use responsibly. I do not condone any illegal or immoral usage of this LoRA.
Description
v1 - Trained on 160ish iamges.
FAQ
Comments (3)
Thanks for this LoRA for Z-Image Base! I saw your post on reddit.com, and commented as "aastle".
You're welcome! This model is really easy to train for, I'm glad I can share it with everyone. :)
Are all the images supposed to looked stretched, like the aspect ratio is off?






