1. Posts must be related to the discussion of digital piracy
2. Don't request invites, trade, sell, or self-promote
3. Don't request or link to specific pirated titles, including DMs
4. Don't submit low-quality posts, be entitled, or harass others
📜 c/Piracy Wiki (Community Edition):
💰 Please help cover server costs.
 |
 |
|---|---|
| Ko-fi | Liberapay |
In addition to the online platforms linked by the other commenters, it's also pretty straightforward to run Stable Diffusion locally, if your hardware is beefy enough: https://github.com/AUTOMATIC1111/stable-diffusion-webui
Various fine-tunes checkpoints for different content and art styles can be downloaded at civitai.
Best way right here. Free, open sourced, and you wont get judged for your outputs.
absolutely this. I have been messing around with this for about a week now.
super fun and easy to set up. I used this since I wanted a docker env.
Prompt: sad monkey working at a computer desk, realistic, office setting Negative prompt: digital art, happy, smiling Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 4, Seed: 509018188, Size: 512x512, Model hash: 4199bcdd14, Model: revAnimated_v122, Version: v1.3.0
Time taken: 13.62s
Torch active/reserved: 2214/2992 MiB, Sys VRAM: 4145/8192 MiB (50.6%)
(side note, Does anyone know why I can't upload pictures directly from web? getting
SyntaxError: JSON.parse: unexpected character at line 1 column 1 of the JSON data )
edit: its because of size....
This is the way. The really top-tier AI art is almost guaranteed to use this, most online tools and other frontends just don't have the features. Also, here is a link to a fork of that with an improved UI (no other changes).
Beefy can mean things to different people too. I have a mobile 1660ti and it can generate images in decent times (about 40seconds for a 20iteration image from prompt)
I'm slightly lacking in VRAM though, something 8GB VRAM would allow you to use most models.
Fun fact, it can be run on as low as 2gb vram! It works out of the box with the --lowvram parameter, and with some extra fiddling with extensions you can even generate high resolution stuff.
yeah thats fair enough on the wordage.
Im rocking a 3070 and 11th gen i7, but only 16gb of ram.
still pretty quick imo
Prompt: Gandalf the Grey dressed riding an ostrich, epic fantasy
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 4.5, Seed: 1771905785, Size: 512x512, Model hash: 4199bcdd14, Model: revAnimated_v122, Version: v1.3.0
Time taken: 3.08s
Took longer for my browser to download the image than it took for you to generate it. :)
Stable Diffusion Web U by Automatic1111 is pretty excellent for doing image generation on your own hardware.
And even if your pc can't run it there are always cloud hosted installs
Didn't google block the usage of Stable diffusion on colab?
Awe yea seems like it :/ that sucks.
Damn. Thought they unblocked it again, maybe ;)
I develop the ai horde myself! It's the best and only one without a catch
what do you mean how do you do that?
Stable diffusion is free, you can install it on your computer
Bing works well up to a hundred images a week then just make a new account if you run out.
you can run SD yourself, super easy to do. Always a guide available on /g/.
Oh I forgot that forum still exists
Yes. Various, with various limitations.
The simplest is Stable Horde/AI horde. Volunteers donate GPU time to the public. Please do not abuse this trust by overusing it without giving back! Great for quick experiments, though, since it requires no sign-up. Also includes a "get probable description from uploaded image" function, I’ve just noticed:
https://aqualxx.github.io/stable-ui/about
Personally, I’ve been playing around with leonardo.ai. It has a payment structure, but the free tier is very generous, in my opinion. You get 150 free daily token. Images cost between 1-4 token. It has a prompt generator, and you can even train your own model for free. There are also a lot of community models, since every model is set to public by default. You can even browse public images and directly copy all settings/prompts into the generator or use them for image-to-image stuff.
The company is very much focussing on building an active community, which can be both good and bad. You are first put on a waiting list, but your account is automatically activated if you join the discord and write some comments. There are also constant community contests/challenges to earn more token.
Edit: leonardo.ai techncally allows nsfw generations, but heavily discourages them. It has an automatic filter for any terms deemed nsfw. You may also not discuss how to circumvent that filter in the discord. (You may discuss how to bypass the filter to prompt safe for work images. E.g. you may discuss how to create Charles Dickens characters despite Dickens being filtered out.)
Here are few more, but I haven’t tried any of them:
https://stable-diffusion-art.com/free-ai-image-generator-sites/
.g. you may discuss how to create Charles Dickens characters despite Dickens being filtered out.)
Ah so the ol Scuthorpe problem.
Exactly. Difference is, you may also create fully NSFW models or pictures, if you figure out how. There’s a topless woman in the top community creations right now. They just don’t seem to want straight up porn, and this makes it difficult enough to keep it from being flooded by that.
I've recently used AI Image Generator. I've also bookmarked this to have a look later as that seemed to be a nice overview of options to selfhost.
You're looking for Openart.ai 😏
Why the fuck is Discord being forced down my throat when I want to use openart.ai or midjourney?
Their business probably gets some kind of kickback for people signing up using their link.
I've been playing around with Dreamerland for the last week or so. The app itself has almost no embedded ads which makes for a clean and usable interface, but it forces you to watch an ad every time you hit "generate" which can get tedious.
https://dbzer0.itch.io/lucid-creations
Is this kind of what you are looking for? It's made by the same person behind this instance
Out of the available image generators, Stable Diffusion is the best but be prepared to mess around with it to get the best results. If you have an iPhone or Mac, Draw Things is a great way to get started. If you want to dive head first into stable diffusion and have good hardware with lots of VRAM then try Automatic1111 as others have suggested.
Yes if you have a spare PC with a fairly recent nvidia GPU and decent internet speeds or a couple hours to spare for inital setup, you can selfhost InvokeAI. You can use stable diffusion with quite a few open models and its not complex to setup.
and
https://huggingface.co/spaces/runwayml/stable-diffusion-v1-5
don't even need registration, but don't expect too much of them.
All those free services I tried, always produce garbaggio results. Where is the actual good stuff?
You can get results as good or better than Midjourney with stable diffusion....but it's not as streamlined as Midjourney is. you have to find the right model on Civitai.com, you have to craft the prompt right, you can use controlnet. And then if you want to go the extra mile, refine it with img2img.
The extreme configurability that comes from running Stable Diffusion locally can be blessing and/or curse. You need to fiddle around and get experienced with the settings to get good results, but once you get good at that you can get really good results.
Not sure how you mean. Stable Diffusion is the golden standard right now and it's free. You just have to run it locally on your own hardware for the full configurability and features. It's a pretty steep learning curve, and requires a decent, discreet GPU.
Stable diffusion + Automatic1111 + custom models and LoRAs. Completely free, only takes an hour or so to set up, and is top of the line.
large models are actually really expensive to run: either run it yourself, or put up with the fact that someone’s not going to pay a lot for people to just use because reasons?
Commenting cuz also curious
I have used https://neural.love/ai-art-generator
Don't have to make an account, but if you do it saves your generated images. Free use is limited to 25 steps, and 512x640 pixels. Paid use expands that.
One thing I've always wanted to do is hack into OpenAI and leak the weights for GPT4 and DallE. Never got around to developing the skillset though (I've done a few CTFs, I'm super bad at them)
Good luck with that. Do good. I can't even imagine what would happen if chat histories were to fully leak
lmao I'd never leak the chat histories, just the weights of the neural network. I have morals, they just do not coincide with the law
If you want an easy to get into sd... check out easy diffusion. The beta allows for lara models now.