Increases details on things out of the prompt, add textures objects and small details. Can be used from -2 to 2, each video and prompt is different so keep it in mind. Trained using rank 1 and skipping the initial layers to avoid changing composition to much.
Description
Initial version -2 to 2
FAQ
Comments (7)
Nice. Finally a minimax H3 Lora that does what it is supposed to do without side effects...
work best on higher resolution, lower resolution (480p) not very effective.
Is this text to video only?
This is seriously good stuff. If there's a way to improve this further, I'd love to see it!
This is the BEST Lora I've ever used for general quality improvement. This lora is a lot more than just detail, it also preserves faces even through scene cuts, and enhances overall clarity of the scene. I can't find any side effects or clashing with other loras. I'm using i2v 720p. Brilliant job here.
Thank you so much for your kind words.
I realized that when I set strength to a negative (-2, -1.5) , I got results similar to the quality of camera phones in the 2010s.
Setting strength to 1 or 2 will create a very "AI plastic" video. However, if it's an animated, anime or 3D video, it's great.