Edit: 1500 steps seems more "compatible" with other checkpoint like smooth mix
Compare both lora choose the one you prefer
Another one yes, but in i2v, it's not perfect  but i have good result in some situations
I tried some new concept training like twerk while riding a bike
Try it and post your  results !
exemple prompt;
Tw3rk, woman presenting her ????????, twerking, shaking her ???, jiggling her buttocks,
Like other lora of this type, ???? is missing :/
I saw a new post recently with a low model lora detailing ????, my advice is to add this one on the low model if you want ???? with a better anatomy
https://civitai.com/models/2143379/wan-22-i2v-edible-anuses?modelVersionId=2424386
You can also use a trick when the lower body is not visible detailing in the prompt "she wears a ?????" Better than no hole :D


Results are better with a Q8 models 20 steps 10 steps high cfg 3.5 no lightx lora 10 steps low with lightx cgf 1
Perfect combo is a 6 sec of 24 fps = 144 frames or a 5 sec 121 frames
twerks at 24fps is just a must have
Workflow are in the video like always


Again do not hesitate to give me feedback

Have fun!