Art style is changing
I'm using wan2.2-rapid-mega-aio-nsfw-v9.safetensors model, I am really bad at prompting I tried the original Wan2.2 but this was painfuly slow with my 4080 but this model is lot faster, but anyway to my question I have an anime picture but when the video finished it's changes to super realistic girl with different colored eyes, I have no idea why as I told I'm not good at prompting I put anime style and that didn't help any suggestions?
I too asked if it was possible to make a model that worked with the wanfuncontrol to video node...and the only way to work on anime...
Lots of the NSFW LORAs are trained on realistic datasets. You might have better luck using the non-NSFW version and adding in anime-specific LORAs for your use case. Also consider using "first to last frame" which may do a better job maintaining consistency.
That's a thing I not doing nothing special NSFW yet, I have a picture a POV short standing and I want her to squeeze her breasts and then called the viewer over, I am really bad at prompting videos I don't know how they work, I do sentences and bullet point nothing is actually doing anything I used different words like beckoning, motioning to approach, waving someone over, signaling to come closer or inviting with a hand gesture nothing is doing it, I did use first to last frame, I generated the first frame the girl squeezing her breasts and the last frame calling you over (beckoning) and that didn't work the video ended before the girl could do anything, and yes I made a longer video but that still done the same thing, I don't know what I'm doing wrong I'm capable making a good pictures but I'm crap at making a videos