Any tips and help in finding apps / websites that lets you create NSFW content with A.I.?

[BernaPuss]

I wish to begin doing a little video editing to create my own NSFW visual novels. Problem is, most A.I. websites I come across do not let you use the A.I. for NSFW content. Also, a lot of words are banned from use.

I need to find a website / or app which allows you to create NSFW content with A.I. (video and/or pics) in order to bring my imagination to life with the appropriate visuals and have as less censorship as possible.

Any tips on where to search?

 

[melantha]

i have very limited exposure to AI-generation websites, but from the ones I know, I'd recommend the ole trusty civitai.
i know it does allow NSFW, and there's a way to get free credits.
but even normal generation is cheap enough i think.

the main reason why i suggest it, is because of the ecosystem.
there's lots of checkpoints, loras, which are also relevant once you decide to go use your own machine for generation down the line.

other sites like yodayo or novelai have either curated models or proprietary ones.
which you can't use or have difficulty using once you use programs like SD forge or comfyui.

on civitai, the checkpoints and loras you'll use on their web generation interface are the same ones you'll download to gen on your own computer. so there's not alot of adjusting when you make the transition. you'll get to familiarize yourself with ponies, illustrious, character loras, etc.

dont value my input too highly though, I mainly use stable diffusion forge in my computer, so website-use is very limited.

 

[mibc9394]

As what melantha said, civitai is good. The idea is to run the ai locally on your machine so that you are not restricted by anything. Stable diffusion webui or comfy ui are some tools that allow you to set up and run different ai models easily for image generation. After installing one of them, you just need to go to civitai to find the checkpoints you want to use for image generation.

Checkpoint is the main ai model used for image generation. Lora is something you use to alter the result from the checkpoint. On top of those, you can also use control net and a bunch of other stuff to fine tune the result. Comfyui or stable diffusion webui have all these ready for you. You just need to download those checkpoints, loras and stuff from civitai, hugging face, or somewhere on the internet.

I think most of the websites online are just wrappers of existing models in civitai btw

 

[osanaiko]

Do you have access to any GPU hardware?
Even a 8gb VRAM GPU like a RTX2060/3060, or even better a 12gb 3060, is enough to run some models locally. and then you have no restrictions at all, except your own skill.
It's fine to do image generation at lower res 512x512 and then upscale using a different service once you are satisfied

 

[BernaPuss]

Do you have access to any GPU hardware?
Even a 8gb VRAM GPU like a RTX2060/3060, or even better a 12gb 3060, is enough to run some models locally. and then you have no restrictions at all, except your own skill.
It's fine to do image generation at lower res 512x512 and then upscale using a different service once you are satisfied

I have a 4070 Ti Super and recently got myself this hyped AMD Radeon CPU which was apparently lightning in a bottle (the AMD Ryzen 7 7800X3d).

Problem is I have no idea how to make the 'homemade' version of A.I. pic/video gen work effectively. I watched a couple of tutorials on YouTube a couple of years ago, but I couldn't make good pictures to save my life.

 

[osanaiko]

Well... that's more than sufficient hardware to work at home. The CPU is cool for gaming etc but has very little effect on running diffusion models. 4070ti is 12GB and will be fine for doing image gen.

The question about how to use a diffusion image gen tool work to achieve the output you imagine will the same regardless if you use online tools or local tools. It's a matter of spending the time to learn how the tools work by playing with them, and researching when you encounter a question you don't understand. "Active learning", as the educators like to call it.

It is more than a little confusing at first, and the field is also changing rapidly so it's common to find a tutorial that is out of date or perhaps the package they reference has been long superseded.

I'd suggest finding a fairly recent beginners tutorial youtube for Comfy UI, and just follow it through the install process and generate your first images.

Choose a model from Civitai depending on what you want to generate (i.e. SDXL for general, PonyXL for more anime style / sex scenes, "Anythingv3" for realistic humans - but there are literally 1000s of models to choose from once you start exploring)

Prompting is a bit of a black art, but once you understand how and why it works (short phrases trigger different image "concepts" in the model, and the combination of these is applied to random static repeatedly until the output image "locks in") you can generally get closer to what you want.

Another way to improve at prompting is to look at what prompts were used on generated images that you like - the user uploaded galleries on Civitai is good for this.

As you start to understand more you can look at other aspects like LORA or img2img etc.