This is my WAN 2.2 T2V Jujutsu Kaisen style LoRA
The Trigger is "In the style of japanese anime"
The LoRA was trained using a contrastive training method, using positive and negative examples to reach the desired style, it is my first experiment with this, so i would appreciate your Feedback.
The LoRA does work with distilled LoRAs (like lightx2v), but the quality, especially of the anime type of motion, is severely degraded, so i recommend using it without them, possibly at a lower resolution and then upscaling using distilled models.
Oh, and you can add "realism, cgi" in the negative prompt to push it away from generating more 3d type of style.
Thats it, enjoy :)
Description
Version 1
FAQ
Comments (6)
This looks great. I wonder why more people aren't posting stuff with it. I think it might encourage people if you posted example prompts to build on.
First of all thank you for the Feedback. I actually thought I uploaded the examples with prompts but now i see that somehow it didnt work correctly, thanks for pointing it out :)
And of course, if you generated some video using the LoRA, i would love for you to share them here ;)
can you do the csm reze arc movie? there's a ton of official trailers for material now or screencaps u could use
Does this lora also work with still images?
i havent tried but i dont see why it shouldnt
the camera angles of this lora is top knotch. The best I've seen so far.