Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Welcome to rpgcodex.net, a site dedicated to discussing computer based role-playing games in a free and open fashion. We're less strict than other forums, but please refer to the rules.
"This message is awaiting moderator approval": All new users must pass through our moderation queue before they will be able to post normally. Until your account has "passed" your posts will only be visible to yourself (and moderators) until they are approved. Give us a week to get around to approving / deleting / ignoring your mundane opinion on crap before hassling us about it. Once you have passed the moderation period (think of it as a test), you will be able to post normally, just like all the other retards.
I see Darkozric says, he was just messing with me. Whateva.
I haven't seen any AI art that makes me want to fanboy it. I look at it, and forget it. Good art has messages put in it by an intelligent artist with life experience.
At the moment people in general, not just artists, fear AI, because they think it will take their jerbs, and there will be no safety net for them. They are right. It will take their jerbs. They better make plans for that.
Mr. Butter thinks that some of us are here to "fanboy" the AI, it's too early to "fanboy" this shit. I'm mostly interested in this as a supportive tool, especially in music. He has a tendency to generalize things, who are those people who afraid to lose their jobs?
If he's talking about factory workers, sure. Personally I have no fear about this since my main job is goldsmith, and I will probably be dead until the AI is capable to replace me.
Also someone should tell him to build defenses outside of his house, terminators are the ones who will destroy him, not me.
I honestly skipped over his sentence about the license "not being compatible with open source" (even though it is and plenty of Debian-compatible licenses restrict use for things like "hate speech")
You also didn't read the links i gave you, or at least didn't try to understand why i wrote you misunderstood what i wrote - instead you thought i am being a dick to you and stopped on that. That approach does require less mental effort from your side, but doesn't help you understand where you were wrong - which is what i expected you to do and why i wrote the discussion would be a waste of time.
But to make it clear for anyone else (and hopefully you too, if you care) who might wonder why these licenses are incompatible with open source and the DFSG...
From the Debian guidelines:
No Discrimination Against Fields of Endeavor
The license must not restrict anyone from making use of the program in a specific field of endeavor. For example, it may not restrict the program from being used in a business, or from being used for genetic research.
Use-based restrictions. The restrictions set forth in Attachment A are considered Use-based restrictions. Therefore You cannot use the Model and the Derivatives of the Model for the specified restricted uses.
[...]
Attachment A
Use Restrictions
You agree not to use the Model or Derivatives of the Model:
[...]
- To provide medical advice and medical results interpretation
[...]
These two are in clear conflict. In fact the entire "Attachment A" is in conflict with FOSS licenses in general. E.g. the Free Software Foundation puts the "freedom to run the program as you wish, for any purpose" as the most important freedom for their Free Software definition (on which all other Open Source and Free Software licenses are based on). The Open Source Initiative also has explicit entries for stuff like that in their FAQ:
Can I restrict how people use an Open Source licensed program?
No. The freedom to use the program for any purpose is part of the Open Source Definition. Open source licenses do not discriminate against fields of endeavor.
Can I stop "evil people" from using my program?
No. The Open Source Definition specifies that Open Source licenses may not discriminate against persons or groups. Giving everyone freedom means giving evil people freedom, too.
Obviously a few things mentioned in the license are about unlawful uses, but that makes them pointless to put in a license - the law would already override the license and people who would ignore the law would also ignore the license.
I don't see what "medical advice" you're going to get from badly drawn photos of women. This license was probably meant more for GPT and meant to deal with legal restrictions on AI in the works. It's there to limit their liability.
I would like as few restrictions as possible on the model. But at this point, I don't see how you are being limited at all by it, so I don't see why you care other than just general anal retentiveness about licenses.
Again, do you use any software that has a license not compatible with the DFSG? You do. So what's the point of this?
Obviously a few things mentioned in the license are about unlawful uses, but that makes them pointless to put in a license - the law would already override the license and people who would ignore the law would also ignore the license.
I agree, but it seems a relatively minor issue for you and me for the time being.
In your first post about it, you just said 'some weird license you have to agree to' (paraphrasing), implying you hadn't read it and mainly objected to the act of agreement because it was strange. The whole reason I responded to you was to allay your fears, not have some long ass argument or to prove you wrong.
Installed SD from the Repo linked in the Guide here: https://rentry.org/GUItard and been fucking around with it a bit too, giving that 3080Ti something to do. Generation of single images is surprisingly quick at around 4-10 seconds per depending on complexity.
After the first few Test attempts of bizarre or relatively hilarious if even somewhat intriguing gens I got some interesting results with simple prompts like "Tower of Babel Cyberpunk Sci-fi Dystopia", "Daniel Dociu Sci-Fi landscape" or "Syd Mead Blade Runner landscape" early on:
I've got some of the most successful results from trying to combine celebrities that likely had a lot of Training data due to large amounts of current photos Online and distinct faces like Ana de Armas or Scarlett Johannson partly based on this thread along with famous artists with distinct and unique art styles like Luis Royo, Frank Frazetta or Kentaro Miura with their own wide portfolio of art Online that will likely have also served as Training data for the AI (which might be a bit of a problem regarding Copyright/Ownership, because it can essentially create untold new images based on the body of work of a specific artist) and a clear theme like "Fantasy", even though it still fucks up the faces in some minor cases and depending on the person used it's nothing compared to some weaker concepts like trying to combine different objects and typing something like "Woman riding a horse" or similar, in which case it'll prolly fuck up the face, body proportions and how she sits on the horse:
There also seems to be a lot of Emma Watson stuff out there for some reason:
On the other hand if you try to input more obscure people or artists the AI has likely not stumbled upon during its Training process you'll have more issues getting anything meaningful out of it.
I haven't been able to generate any impressive looking Eldritch abominations (other than the unwanted ones at the beginning) or complicated Aztec design, character art and landscape stuff like some seem to be able to get out of it that I've seen Online with long elaborate prompts and prompt refining/iteration so far though:
What I haven't played around with yet and almost seems more impressive or powerful is the img2img part of it, people can quickly throw together the most basic and shitty sketch imaginable in MS Paint or on a piece of paper, input it into the tool and along with a prompt of what they want it to be or become get impressive results based on their composition:
Similarly existing images can quickly be replicated into a (or many) different styles using this:
Overall it seems like a very impressive and powerful art tool that is only going to get more-so in the coming years as it grows up out of the early stage of infancy and gets more competition if they don't intentionally fuck it up for ideological reasons like some of them are trying to do:
That Twitter convo is pretty funny when you realize that chicken little there works for OpenAI, Stable Diffusion's direct competitor, and that they're the ones who released the CLIP model Stable Diffusion was based on in the first place.
I see Darkozric says, he was just messing with me. Whateva.
I haven't seen any AI art that makes me want to fanboy it. I look at it, and forget it. Good art has messages put in it by an intelligent artist with life experience.
At the moment people in general, not just artists, fear AI, because they think it will take their jerbs, and there will be no safety net for them. They are right. It will take their jerbs. They better make plans for that.
I agree about the imagery. Won't amount to much until a master artist or someone with a deep knowledge gets their hands on it. Right now it's mostly "gamer" type shit and pedestrian fine art bullshit. The gamer stuff is pretty good already, but that's not the highest bar to clear since a lot of game designs are nonsensical or dumb and are like assembly line product.
As for the other thing, I'd say the problem was relying on A job in the first place. Multiple revenue streams has been the meta for a long time now. Entrepreneurship is ideal, but everyone is not cut out for that. Job security is an illusion, or it should be treated it that way. We aren't entitled to anything in this universe, least of all having fun at work or doing something we're passionate about, which is why a lot of people go into art professionally. Art, the paid kind, used to be a trade, it was work, not play time. E. Michael Jones calls art "a refinement of labor". I like that. There was practical demand for the skills, I don't think their heart's desire was to do church pictures. Trying to bring something out of a block of marble is not for green haired soy sisses on Twitter, it's manual labor. But art has been a luxury for a long time, and luxuries are expendable.
I've studied this AI generated phenomenon quite a bit lately, and I find it fascinating.
I've start to harbor an idea that if the original skill level of the artist is of moderate quality, the end product, be it assisted by AI or not, might amount to very little. Whereas in contrast, a gifted artist may see his or her skills even further enhanced by this wonderful invention, even if it's still in early stage.
A case study:
A Portrait from Numenera and its AI enhanced counterpart:
****
The same portrait of Erritis from Numenera Wild Edition (Also known as Masterpiece Edition) and its AI enhanced version
****
See, it's like night and day?
Granted, the resolution of the originals is different between TToN regular and Wild Editions, and I may have used a slightly different prompt, so this might not be an experiment conforming to the most rigorous scientific standards, but to me it clearly survives the practical test -- what would you rather see in your game.
Yeah, i still don't care about proving you wrong to yourself because you are either unwilling or incapable of realizing it, however since you tried to write to someone else that something i wrote was wrong, i wanted to make it clear that this wasn't the case - we are in a public forum after all, i am not sending you DMs that you only can read.
As i already wrote above, having arguments with someone who is more interested in "not being wrong, any means possible" than correcting themselves is a waste of time, i've been discussing stuff on the internet for almost three decades and experience says that people who do not accept they were wrong after the first reply or two, pretty much never accept their mistake and instead go on making all sorts of irrelevant and sometimes illogical replies (regardless of how they relevant may look at a surface level).
I don't see what "medical advice" you're going to get from badly drawn photos of women. This license was probably meant more for GPT and meant to deal with legal restrictions on AI in the works. It's there to limit their liability.
...here you use some specific interpretation of a specific use case that the data might have in an argument about one license applying to some guidelines for licenses or not - this is completely and 100% irrelevant to what i wrote and it only serves as a way to sidetrack the discussion to something different (the specific application of getting medical advice instead of if having that restriction in the first place makes it incompatible with the guidelines).
In your first post about it, you just said 'some weird license you have to agree to' (paraphrasing), implying you hadn't read it and mainly objected to the act of agreement because it was strange.
This is you completely misunderstanding what i wrote and why i wrote it, which is why in my response above i told you to re-read what i wrote and get rid of any existing misconceptions you may have.
Now if you want to assume any post of me here is a continuation of the discussion, regardless of how false that is, then i can't really stop your ego stroking - after all if i was able to do that you'd actually read and understand what i write.
Ah, I see. TBH, I stopped following Numa numa at all once I say they were going to do a non-white love interest. I viewed it as a bad sign and another departure from the formula they were funded to recreate. Looks like it was.
reminder that the numanuma owners required them to make all the characters ugly after they found out there were *gasp* wypipo -- the horror -- in the game
And this is another reason why AI has potential in gaming. Anything to lower costs and increase the agency of indies so that they can bypass this industry filled with literal cuckholds in as many ways as possible.
one of the more interesting things I see in these images is when it attempts to include language, it's the most alien looking characters I've ever seen
Yeah. That's because the AI has no idea what text is, and since it varies so often with different fonts and languages, it doesn't really know how to draw it.
It's kind of similar to the problem it has with hands.
Assuming they aren't a celebrity or famous character, the easiest way is to use img2img with masks or feathered selection in Krita to alter their expression.
If you need to generate, try using the same seed and adding keywords and/or using negative keywords, but it will probably be hit or miss and change other things, if it even works at all. Which is why img2img is the most precise, if you want small changes to an image.
If you're really determined to generate the same character in different situations/expressions, you might also try using textual inversion on a handful of images of the character. Then, hopefully, it will generate their face, even if the seed and the prompt are different, though again, this will be very hit or miss.
There are ways, either prompt editing or what's called "img2img alternative test" in the Automatic1111 build. I tried the other one and it seems to be a bit buggy so far, the images produced contain noise and the colors have incorrectly increased contrast, but I assume that's temporary.
The way it works is that it tries to reverse the diffusion process and find a seed which, together with using an accurate prompt, most closely resembles the input image. The "interrogate" feature helps here, it tries to find the most accurate prompt automatically. Then if you change a few words it generates a very similar image with only subtle changes, works for different facial expressions, different hair, etc. Sometimes it breaks and suddenly creates something completely different.
It's very practical: I found that women are quite happy if you send them their portrait processed with words like "middle-aged, old, unattractive, wrinkled".
It's hilarious. Every time I hear some talk about the future of AI and it's impact on society, especially at the workplace, I'm hearing stuff like "this will first and foremost affect highly repetitive jobs, either replacing them or massively changing how they work".
This license was probably meant more for GPT and meant to deal with legal restrictions on AI in the works. It's there to limit their liability.
