LoRA works with WAN2.1 image-to-video and text-to-video.
Keyword: fuck you gesture
Description
WAN2.2 Update – Middle Finger LoRA
This update brings improved performance and flexibility. The model works in both I2V and T2V modes.
Keyword:fuck you gesture, a hand showing the middle finger, palm facing the viewer, rude and insulting gesture, close-up on the hand
WAN2.1
LoRA works with WAN2.1 image-to-video and text-to-video.
Keyword: fuck you gesture
FAQ
Comments (13)
Amazing one! It can make things very spicy :)
Nice! I was just thinking I wanting just this lora and here it is! will check it out!
thanks finally
This will be fun lol!
I was surprised how difficult it is to get Wan to do the birdie.
ai struggling with hands?
nahhhhhh
great idea
Not always the middle finger, but works great.
Format is unusable for me (doesn't match other formats). I'm guessing this isn't a Wan 2.1?
It is indeed the wan2.1 model. Could you give me the full error from the log?
I basically get thousands of lines of the following during the conversion step:
WARNING:__main__:unexpected key: lora_unet_blocks_0_cross_attn_k.alpha in diffusers format
WARNING:__main__:unexpected key: lora_unet_blocks_0_cross_attn_k.lora_down.weight in diffusers format
WARNING:__main__:unexpected key: lora_unet_blocks_0_cross_attn_k.lora_up.weight in diffusers format
So... I'm kind of at a loss.
AlbanianDoorknob Can you try it in t2v? I trained it with this model, but it should also work well with the i2v model. You can also try changing the LoRA loader, like Power LoRA or the one from Kijai
AlbanianDoorknob "I’m currently preparing the same thing for the WAN 2.2 14B and 5B models. If this doesn’t work, I hope those will.
Lora addictif !!!! je ne me contrôle plus !!! je FuckYoutise toute mes images !!!!