text2vid img2vid H3 Minimax - Fully uncensored local AI video generator released today

Sorry no goon in this vid yet since I'm working my way through the chapter. A total pita battling with dialogue for the past two days generating this thing. Pro tip; the (S1) and (S2) tags for dialogue are used to tell the model which subject speaks first and who speaks second. Was using these wrong for quite a while and it resulted in my audio references getting totally messed up as the model kept assigning them to the wrong female. Anyways I'll eventually get to the sex part from this chapter

 
anyone had any luck with breast and nipple shape/size/color references in ref2va? I feel like H3 is always making generic looking breasts even when provided with reference images and prompts that refer to them.
 
hey for your videos on your minimax prompts thread, what MP do you use or recommend and settings wise? I already use only 12 steps with er_sde/beta with no lora. Your videos look really clean and sharp.
 
Does someone noticed with rendering time huge differences with exact same prompts and settings? One time 5s vids need ~1,5-2mins and another time 6-8 or more minutes? Usually the results are even better (usually a lot!) than wan 2.2, but its weird this big differences. wan21 or 2.2. never got this.
 
You must log in or register to reply here.
Back
Top