Search results for query: *

What the fuuuuuuuuuuuuuuuuu-

Maybe I'm being a bit to harsh though cause this is the first time AI has been used in such a large professional setting, maybe WIT doesn't have the hang of it yet.

Yes but its sad and disappointing and somewhat hilarious that redditors with 0 art skills in their basements can outdo an entire anime studio.

Cause recently a person making a comic with AI got there copyright for it revoked.

You should probably go to a copyright office to ask about the legality of that first before selling, if I remember correctly you can do it online too. Here: https://www.copyright.gov

Yeah. Like its fine to just use ai if your using SD as a hobby but in professional work? I expect some level of actual art to be done to at least coordinate the AI Something like this. Not this bull.

Also- apparently they fired a hundred people for this... Goddamnit Guess it's expected when you have something like this people would start cutting corners...

Like its seriously not that hard to make a consistent good looking background with SD. WIT didn't even try

Okay You know what? Let ai stay in the bin a little bit longer.

I realized that after I posted it:mjgrin

I talked to him, he isn't being sarcastic at all, he actually wants people to perform copyright infringement. He thinks it'll somehow get MJ into trouble Also yes I agree that creating a "Mugs Money" and commercializing it, would be infringement. But from what I've read, if you commercialized it. I'm not really sure of this so I could be wrong, but from what I've read, if you aren't making any sort of commercial value, you are protected by Fair Use. Which is how fanart and fanfiction is protected. If you started selling Mugs Money then you can be sued. Idk about MJ but SD uses a license that allows for commercial and non-commercial usage. Meaning that if you generated a copyrighted character and simply shared it on the internet for free you'll be protected by fair use, but if you sell that image on a t-shirt, you can be sued. At least that's what I've read, I could be wrong. EDIT: Though you probably already know this.

MidJourney is kinda fucky though cause you do need to pay to use and the copyright is weird, so dude might have a point but still doing this would put the person in trouble too. Good thing I don't use midjourney.:jeff Oh them fuckers, I agree they give the pro-ai crowd a bad rap. There are bad eggs in every community. Maybe it's cause of perspective but from what I've seen most of the toxicity came from the anti ai crowd. But twitter is a cesspool so it doesn't surprise me that both sides are toxic.

I'm not talking about all artists, but those twitters artists who purposefully spread misinformation and toxicity? Yes fuck'em. Look I get being angry at ai cause it means that people don't have to pay you draw porn anymore. But when that means -inciting people to perform copyright infringement by generating images of characters and then tell people to put them on products and sell them. -demonizing people who use ai to the point of wanting them to die or calling them Nazis -and spreading misinformation without actually doing your own research and inciting other people to attack ai users. They've passed multiple lines and I'm going to insult them over it.

But like truth be told. This paper is not an actual research into Stable Diffusion and other gan models. It was definitely made for the Anti-ai crowd so they could keep spreading misinformation about how ai steals images, while anyone with a basic knowledge of how SD works and reading comprehension could understand that the results were extremely biased. But luddites don't care to do any research into how these stuff work anyway and not listen to anyone that doesn't conform to their opinion. An actual unbiased way to see if the ai can reproduce images would be to use a base model (not a trained one) with some non-biased prompts and then generating images to see if the ai can generate a copy with barely any human input instead of what the creators of the paper here did.

Except it isn't really imagining it? 1. It wasn't random, the scientists went through multiple generations to make the images, while explicitly skewing the ai into generating these copies, getting the results they wanted. 2. I see it like how normal copyright infringement works, i.e you drew is copyrighted work, so you can be sued but the actual program used wouldn't. I don't think there's a case for suing the ai for "memorization" either, cause the ai doesn't save any images, and the chances of the ai producing a perfect copy of a copyrighted image in a single generation without any sort of skewing is less than a million, the paper itself has said that the chance is extremely low. Only way to get such perfect copies would be by purposefully skewing the ai into generating them, like what the paper said they did. To the point where they generated million images and didn't find copies.

Point is that the results are extremely cherrypicked. First off they forced the ai into producing copies by generating millions of images, taking the most similar ones and generating based of that, hundreds of times. Obviously if you are pushing the ai into making a single result eventually you'll get it. But the ai randomly generating a copy is a less than a million chance normally (this is stated in the paper itself) and it took millions of generations to actually get said copies. Two the second part with making their own models is a classic case of overfitting, where a single image repeats in the dataset so many times the AI actually does start copying it. It should be noted though that this is a problem. I myself ran into this problem when training my own model. Here is the Huggingface page for it where you can see the four images I used to Dreambooth train SD provided by @Nevermind https://huggingface.co/sd-dreambooth-library/oldcybrpunk When I first trained the model I...

https://arxiv.org/abs/2301.13188 Lol, twitter artists tryna say gotcha with this paper when

https://imagen.research.google/video/ Text-to-Video is already being developed, kinda insane how fast tech progresses. I really want to know how they made the AI make consistent frames, cause using Deforum, SD has a hard time keeping things consistent.

My mom may have caught covid unfortunately, we don't know yet though.