I run everything local on my machine and use checkpoints that have been trained with a 'bit' of porn (some have too much and lead to unfavorable results). I also use LoRAs.
I started with an image of a woman with ginormous breasts I got from various sources (local, free web-based image generators, etc.). Pony with a checkpoint trained on real life images instead of art is great for big boobs because the LoRAs trained on anime and shit still work on the life-like images. Then I take that image and plug it in my WAN 2.2 AIO I2V NSFW checkpoint in ComfyUI.
See the attached screenshot of my settings. I have been too deep into this. While my results are not always the best, I can crank out a lot of fairly good stuff. I wish I could afford KlingAI, but I am slowly learning how to work around it. WAN I2V is the next step, and having a piece of software like Honeycam helps for taking the last keyframe of a series of images and feeding it back into WAN to do the next segment of a video. I could make entire movies like that if I wanted to. I have S2V as well, but I don't have any good source for emotional audio. It all sounds flat no matter what I use. I tried some of the "emotive" voice software and I don't like the lack of control of the emotion.
Anyways, I have seen your stuff and you do great work, better than anything I can crank out. I go for the amateur video/camera look that "could" be real if ginormous breasts were possible. Think photos of bigfoot, only with better lighting and focus.
Punchy