I cooked in Veo 3 over the last days. More or less successfully. I only got Pro, not Ultra (actually got 2 Pro subscriptions but anyway), so I am limited in my daily gens and can "only" use Veo 3 fast, not quality, although even the fast mode is impressive. Sadly, even after refining my promps with several guides, getting a 10/10 generation still comes down to luck. 30% to 40% are sorta decent / passable, with 1 in 10 being perfect and making it seem like magic, those are the kinds of clips trending on social media, youtube etc. Sadly, I can't use flow for some advanced features (not available in my country), besides, I hear some stuff like scene extend is a bit buggy / nonfunctional. So, to keep consistent characters, you need a precise, detailed and specific description that you keep reusing in your prompts (and even then you need a bit of luck - the second sequence in the first part is almost perfect, though the girls switched sides, and the blonde suddenly looks a lot thinner...)
Fighting the safety filters is a bitch, you need to sanitize / tone down many things like physical descriptions (i.e. behind instead of ass, chest instead of tits or boobs), and you can't directly type that you want vore content since fetishes are banned. But if you just stick to the physical descriptions, you can try to weasel around the limitations (instead of saying a char has a fat gut because they ate someone, just describe it physically, the model doesn't need the context which would only trigger safety filters). Hopefully a model just as good as Veo 3, but not filtered as fuck will come out sometime (Kling 2.1 apparently is also good, but Veo theoretically is capable of better audio output and you don't need any audio post processing, which I have no clue about whatsover). Oh yeah, celebrities / notable people from real life are also restricted, I tried to make an Emma Watson like girl but I couldn't use the full name. A detailed description plus first name is okay I guess...
This isnt done yet but I feel like I am going mad trying to get this right, 3 days / 18 generations of trying, this is what I have so far. I heavily stitched clips together and cut various parts that had weird things happening, bad audio etc, used CapCut for that, saw a youtuber recommend it. Never done video editing before, so this is rough, but it's something. Final product hopefully better. Had to split this in 2 parts, apparently there is a limit for the size of individual files - and one for the size of total files in a post, since this would be exceeded as well, gonna post part 2 seperately.
Ah, and the last 2 sections in this vid are work in progress shots made with Veo 2, the far inferior model and it has no audio. Since I have some Veo 2 gens each day after I use up the Veo 3 ones, I find it useful to test if my prompts might pass the filter (not 100% guaranteed it will work in Veo 3, since those filters are even stricter imho, but it's a start), and if the AI gets my idea at all. I threw them in to show what would come next and hey, it makes for a semi decent vore AI vid I guess.
Anyone else experimenting with this? I saw those clips from the 9th of June, not bad, but I wanna try to see how far I can take this / what the limits are. Internal shots of a digested prey? Maybe post vore content? Will see.
Hope someone gets a bit of enjoyment out of this, or this inspires some to create stuff, I know Veo 3 is pricy but when it works, oh boy. The first sequence in this video absolutely blew me away (turns out I got lucky there I guess). Anyway, enjoy and have a good one...