Does anyone know of a good AI site for producing pics? I know there are some good ones out there, and I've played around with a few and had mixed results. What I am looking for is a site that will produce pics of a certain person in a certain outfit/pose. It doesn't have to be porn though, it's more about the person and outfit which for what I need can be SFW. Is there a site where I can up load a photo and ask it to produce a pic of that person but in a particular outfit? Or maybe just use the models name (the models I have in mid are porn models though, but as I said the pics I am after can be SFW). If the site is good enough I don't mind paying a small fee/subscription, but if I'm paying the results need to reflect the quality of the pics.
I'm using Pixai and am overall very happy with the quality. You get a generous amount of free points each day and can safe them up too. Only downside is that they have started to clamp down somwehat on NSFW stuff, (which used to be the main appeal of the damn site lol) by banning some prompts outright You can still produce erotic pics though and when that isn't really the focus anyways I think it's a pretty fantastic free option
Do you want to do the following? 1) Input an image of a person 2) Write a prompt like "this person but sitting on a chair and wearing a t-shirt and jeans with sperm on her face" 3) Get what you requested. I doubt that any modern service can do it in a convenient way. AI's do not understand image inputs like that. Neither do they know the names of minor celebrities. You can train a LORA on someone's likeness (simplifying LORA = mini-model that is then attached to the main one) and then AI will start understanding what person you want to get but it is not a single button press process to do. PixAI kinda allow you to train LORAs but... it isn't optimized for photorealism, at all. And as Gambio said, they are moving away from NSFW stuff and may not even like your attempt to train on a porn actress. If the porn actress in question is popular, some kind soul may have already posted the LORA somewhere but for minor ones, chances of finding one are very slim. Then... you need a service that allows you to use LORA's or learn how to run AI locally (or on services that allow remote execution of python code and offer good GPUs) .
You're correct, that is what I want, and you've confirmed my suspicions that it will be difficult to impossible to achieve. (SPOILER ALERT FOR PROJECT HAREM) I'm writing a major scene where 5 women in my story (all of whom are based on IRL porn models) go to a fancy dress ball. I thought it would be nice to have pics of them in said fancy dress and wondered if AI could produce that. So for example, I want a pic of Karlie Montana dressed as Jessica Rabbit. I'll continue to play around with for a while but I don't think I'm going to get what I want. Unless I go the other route and start playing with photoshop or something, but that's a skill I don't possess or have time to learn.
You may try inpainting, basically photoshoping but with an AI - you keep part of the image and AI redraws the rest. There are places online that allow you do it: PixAI, AI horde interfaces, others... I rarely use online services nowadays, since I got a new PC. Having a powerful enough GPU, I can locally run Stable Diffusion with resolutions up to 1024x1024 (even more if sloooowly). Some new services may have appeared but I have a vague idea of what is required to do what you describe and don't think there are few click solutions.
It’s basically a small model trained on one specific thing that is combined with your main image generation model so it’s capable of generating new kinds of images. So if I’m using Stable Diffusion XL as my model checkpoint, it wasn’t trained on images tagged as Karlie Montana, so it won’t recognize the word or generate the image. But if I add a lora trained on Karlie Montana, and type in a specific keyword specified by that lora, the main model will now be able to generate pictures of her. This is all set up in whatever interface you’re using to generate your images.
There are some AI tools that can generate custom images based on a reference photo, but most mainstream ones avoid using specific names or real people for ethical and legal reasons. Your best bet might be something like Leonardo.Ai or Stable Diffusion (with custom models), especially if you're open to tweaking prompts and using inpainting for better results. If you're looking for hyper-realistic edits, some paid services like Fotor or RunwayML might be worth checking out.