>>19769665>"blowjob, deepthroat; The view is POV from above. deep-throat oral sex on a penis. a cute woman performing a deepthroat blowjob. She's naked and have a big breasts. She kneeling in front of a man. The man is moving his penis back and forth in the girl mouth and throat, while she is sucking the man's penis, the penis is going deep into her mouth and throat. She looking at viewer happily."I think that would work ok for T2V. For I2V there are loads of contradictions and/or useless stuff (like pov from above, tits not visible so why prompt it, this will most likely cause cause AI to look for them by moving camera, makes girl trying to undress....aka NOT focusing on the action you want, deepthroating in this case)
Start simple; "Redhead girl kneeling in front of a man with erect penis. blowjob, deepthroat;"
And don't overload with loras. Use one or two at the same time.
Also note that some loras work better with lower CFG
https://civitai.com/models/1515071/wan-21-deepthroat-facefuck-entry-blowjob-fellatio-oral-lora-t2v-14b = good for insertion, use at lower strength, 0.2-0.5
https://civitai.com/models/1497390/deepthroat-wan-21-i2v = the action itself, use at 0.75-1.0
also these
>>19769622 what friren said, but on dimensions, you don't have to be exactly 480 width/height, but make sure your dimension is divide'able by 16......480, 512, 576, 640, 720, etc