I’m a 120v guy myself but I won’t judge
It's in the name, I run on that extra potent Yuro juice.
I normally am super hung ho about new tech but I’ve really not had much of an interest in using these tools often enough to care. Maybe use a free one every great now and then.
But I had a need to use them a few times today and it was just wild to see how pozzed their model had become. Like literally ignoring my entire prompt and generating women in American flag hijabs surrounded by unity themes text it pulled out of nowhere, or just rainbows and rainbows…
Like I get it, they don’t want me making a Nazi cow but….
It's very easy to get started with local genning, and it will give you way more flexibility than any online tools. Whether you want to gen some extreme, niche fetishes or some really specific joke, you'll have a model or a LoRA specifically trained for that purpose. All you'll need is a rudimentary understanding of following readme instructions, file management and configuration file editing, a gaming PC, ideally with an Nvidia GPU so it just works, plenty of free disk space, and plenty of free time.
The current two main players in the WebUI sphere are:
reForged - Yet another A1111 fork that seems to be taking precedence in the A1111 fork world. Gradio based, got everything you need from the get-go, piss easy to use, this will be your starting point in the local genning journey.
ComfyUI - The good ol' node spaghetti UI. Painful to set up at first, especially when trying to replicate more complex workflows that A1111 derivatives come preinstalled with like regional prompting or ControlNet, but it will give you far more flexibility and better understanding of how Stable Diffusion works.
I'd only encourage using Comfy if you already have experience with A1111 and you have the autism to take the next step, or if your hardware is so weak that A1111 shits the bed. Comfy's backend does very well in offloading when you're limited in resources, that's how I could run SDXL models on my 1060. Both are trivial to "install", you just pull the repo, run a Batch file, and it'll pull all the necessary dependencies for you, get everything up and running and send you straight to the WebUI.
Obviously you'll need to download models to be able to do anything, and for that you'll have a few spots.
CivitAI - The Mecca of AI degeneracy, but if you're not into that you'll have plenty of SFW resources for genning memes and other fun stuff that's not meant to get your rocks off.
Hugging Face - Here you'll find more generic models, like the base SDXL/Flux models, base ControlNet models and whatnot, but also trained merges. Still a good secondary source.
If you end up installing two WebUI's at once to switch between them, you'll want to read into changing the model lookup paths so that both can share the same collection. They are hefty, my model folder is over 100GB, so you will want to avoid having two copies of it.
Do keep in mind that you'll have more control over what you're doing outside of a single prompt box, and it may seem overwhelming and complicated at first, but you should be able to quickly get the gist of the basics like what resolution, prompt, seed, CFG, steps, sampler and scheduler options do, how they affect the gen, and then you'll be able to gradually move towards the more advanced techniques like hi res fix, inpainting, LoRA, ControlNet, regional prompting, IPAdapter, ADetailer and so on. Half of the fun in local genning is figuring all of it on your own and experimenting.