This LoRA is designed to give your images a storybook animation fused with concept art sketch style using saturated colors and rough outlines. The trigger "word" is the start of your prompt and explicitly tells the model to use the style, as well as maintain it, since minimax does tend to revert to realism over longer clips. Add whatever you like after the trigger.
Recommended Strength: 1
Trigger Prompt: This video is shot in RSNature style animation with defined yet rough sketch lines, saturated painterly colors, and concept art draft quality. This style remains consistent across movement and scene changes.
The turbo LoRA is much cheaper, but sadly does lower quality some... The showcase videos were made without turbo, at 20 steps. This is admittedly my first video LoRA, but the results actually made me very happy. Please post anything you make with this to the gallery, I would love to see it! Thank you so much for your time and support! 🌷
Special thanks to several friends, and especially @RisingV for some help, advice and insight into training this kind of LoRA that saved me a lot of time, buzz, and headache! ❤️
Description
First Version.
FAQ
Comments (14)
Very beautiful style!
Thank you so, so much!! I had a lot of fun with this (つ｡˃ ᵕ ˂)つ ⸝♡
Thank you for the shoutout. This looks really great. I need to try it myself soon.
Thank you again for help with the images suggestion, and what I picked up overhearing your tests of your lora, and confirming the license question for me... Figuring out all of this stuff in a hurry for the contest submission was crazy, but I got a lot of guidance and assistance, and am so happy to have pulled it off (˶ᵔᗜᵔ˶)ﾉﾞ
@FreijaFoxy No problem. I guess noone of us will win the popular vote judging by the speed Citron is putting out H3 LoRAs (maybe 2nd and 3rd place are still up for grab ;)), so there wasn't that much pressure to get it out fast, but at least we did it.
This gives videos such a nice style and the consistency is amazing, using the suggested prompt makes the style come out clearly every time. 5/5 Stars, highly recommend.
Thank you so much for trying it out! Your videos are amazing, I'm so grateful to have them displayed in my gallery!! 💗
Awesome! Would you be interested in training a LoRA in the same style for image models like Krea 2 or Qwen 2.1?
I have never used Qwen, so that one I really don't have enough experience with, but Krea 2 and even Illustrious might be quite doable, I will add it to my list of things to look into for the future, thank you so much for the nice comment and interest!! 💕
Thank Youuuuuuu this is absolutely amazingly awesome! Incredible quality and user friendly too! Green thumb up!!!
Thank you so much, I'm so glad you're enjoying it!! Your adorable ghosties are the perfect start to October, thank you for sharing them!!
@FreijaFoxy TY!! My pleasure! Ooooooh I love your new stickers!! Magically cool!!!!!!😎
What a fun Lora!
Thank you so much, I really appreciate that. I was truly amazed by how well MiniMax H3 absorbs styles. There's a challenge to include enough training data about diverse subjects that the model can "imagine" ones it hasn't been trained on, and I felt a sense of urgency to finish this in time for entering it into the contest- I'd already lost 5 days researching how to train for this video model and preparing a new project. In the end I feel like, even seeing the spaces between the kind of nature and themes it was shown, it comprehends the rest really well. I wish I could say it was due to amazing skill on my part... but MiniMax H3 is just a very capable model. I sincerely hope someday they open the license terms to allow local use of it for the currently restricted countries, if only I weren't limited to civit only for this I would have loved to explore even more ideas, and still might, someday.