As far as AI goes, these are really good! That said, there are still a lot of visual artifacts/errors with these for me to implement them into my workflow, I'm afraid.
The one with Grace is especially
nearly there, but there are still a lot of, albeit minor, issues with it.
That said if
Cashhh is confident he can work these kinks out (pun intended) and possibly provide longer, more consistent scenes, please send me a message!
PS: The question is also how long these took to make!
I wish I had the time to dive into a project like this, but currently life is trying really hard to kick my ass. The good news is that, anyone with decent hardware and some time on their hands can do this.
I'll clear some things up with my limited knowledge and provide some info for you and the SP community.
The NSFW AI video models are fairly new. The one I'm using in the examples (Wan 2.1) is only two months old. AI is moving at lighting speed, and new stuff comes out every week, but as far as NSFW video, Wan is the go-to model. I'm using a smaller Wan model, that fits my hardware (3080 10GB/32GB Ram). With better hardware and settings (workflow), the results would be artifacts free, and look much better.
The examples I provided take about 20 minutes per try. I call it the AI roulette, since you'll need at least 3-4 tries before you get a decent result. However, with better hardware, the time can be cut in half or more. The most consuming part timewise is acquiring the knowledge
I've been playing around with AI image generation for about a year now, and only a couple of months with AI video.
If I had to do research on the AI video stuff again, I would go to the links below. Don't mind the bad audio quality, this person is very helpful.
Youtube channel:
You must be registered to see the links
Website with settings (workflows):
You must be registered to see the links
For all AI "art", and NSFW helper models (loras), (you need a free account to see the "goods"):
You must be registered to see the links