i think i get what you mean now. so its only possible to have them generated using local uncensored model like ltx or wan for i2v tasks like that,correct? and the major downside is that it is way slower since it's done locally or via cloud computing?