Yes the biggest stopper for AI at the moment is just the shear hardware requirements to run it. That’s why I see it as a service in the near future.
That being said, computers will improve no doubt, the problem is how available will be the hardware.
Personally I’m working on 5-7 year old computer at the moment. Can’t afford new CPU, motherboard or a GPU. Which means that even if it’s free, it still might be out of the reach of many people who cannot afford the hardware for it.
5 Likes
I’ve seen people using it as a starting point and I think that is pretty valid way to use it, but to create art with only AI and no human input other than text is cheating in my opinion and the robots should get the credit rather than the person if that was the case. Some people will probably use it instead of commissioning an artist, but they will end up with generic images that are pretending to be famous concept artists styles or pulled from clip art websites.
sort of related subject- I think AI tools such as de-noisers, create depth map from 2D images, background removal and image enlargement are very useful.
5 Likes
A thing no one thought yet is legal usage to use image to generate others. I am almost certain there will be many people copying other artists and when it hits a big company like copying a style of a big game or movie there will be a limitation to what can be used as in the style of something else. Despite the technology being all fancy it will cause upsets and be limited probably. Or they will just have a human change it enough to not fall on it to by undetected. But it will happen once names like Adode or Autodesk start to do it. But I am sure there will be drama. Why pay a artist when you can mimic all his work because it is all on the internet?
Or maybe you get payed a cent for each time they copy you.
2 Likes
I tried this site out a few times. I have had no issues with the usability of this site. I sign in with Google. And select categories ex; buildings. Chose two images and viola. It uses AI to blend two or more images into one. You can explore images that other users came up with, tweak with it, mix it up. It’s simple enough, free, fun, no dl needed and it’s (CCO). I can see myself using this for inspiration/exploring. But I have no idea what this means for the future of art/humanity.
https://www.artbreeder.com/
3 Likes
I made the mistake of Instagram following some of my favorite artists who use mid journey and that’s not the problem, the problem is I have seen all the artwork I hoped I would get skilled enough to do and hoped I would learn how to do by artists (?) typing in some prompts. Many of these artists are great artists on their own without the tool but I’m not. When I started my art journey a few years ago I basically just looked at art station for some creative juice and now all this stuff I see is what has been in my head with no idea how to accomplish it. I think I will just not think so hard, that’s easy! And I will enjoy the process.
Notice it says “prompt”
4 Likes
Art breeder is a nice tool. It has a different way of working using “genes” to have a design evolve with a bunch of sliders. It is fun to play around with.
People are calling these tools “AI tools”, but the intelligence is really just that a computer is doing work. I think a long time ago image filters like blurring and sharpening were also considered “AI” back in the day as well.
2 Likes
Yes, I think a lot of artists are reposting their image results on things like Instagram or Twitter. I am not sure where the word “prompt” came from, but that seems to be what everything uses for “text I wrote that made this image”
1 Like
So far I’ve found Midjourney useful for finding moods and light/colour references, also I have used it once or twice to generate material for a photobash-type workflows. For me, the lack of real control over viewing angle and composition more generally makes the currently available batch of tools pretty limited. I’m sure that will change, though.
4 Likes
That is kind of my experience with MidJourney. Some of the ideas are neat, but it kind of just uses the exact same compositions which can look generic. It doesn’t do multiple characters very well too, but it is good to get past the blank canvas syndrome. Tyring to get characters interacting just seems to create nightmare illustrations like this
(supposed to be Thanos fighting spiderman)
5 Likes
Here’s an interesting and relevant article:
3 Likes
Thanks for the article. That was an interesting read. It should be interesting to see what happens in this space in the next few years. I listened to another long audio chat from David Holz last week. He was thinking auto-generating 3d models from text is going to be next and not that big of a leap from generating 2d images.
1 Like
I’m on a human emotional level is uneasy about it.
I can see myself maybe using it as tool for inspiration… .oth I can’t remove on the back of my mind other implication. Visible art style by known still alive artist can be feed and art can be produced, for most people thats enough for their need… and then they won’t be paying for the service/royalty of the artist whose artstyle was use as as basis.
There’s a dread i’m feeling about it - this despite me being pretty open to technology in general. Feeling like there is an ethical ground we walking on and the feeling that some portion of the industry [like concept art for movies / will greedily use it - as replacement / or a way to pay artist lower]
I love looking at beautiful art pieces as much as trying to create one.
OTH a random shower thought - but then this is a learning algorithm that learn on art it gets feed it [so it does better that more fresh art its feed], what if it eventually feed loop it with its own generated image. 
- eventually one of its own generated image will get feed back to it.
something something to think about.
2 Likes
Well yes there is an ethical ground. Just like work on DNA and genetic modification.
And yes a lot of high tier artists and studios will exploit the tool like hell.
So far however, although cool looking the AI prompted images quickly becomes noticeable and repetitive.
2 Likes
I once started a Python program - initially even thought of as a plugin - to create lineart based on sketch. It’s one of the things that image-based deep network AI should be really good at, though of course it would be best with huge network, great hardware to train on, and huge database of pictures, while I would have to create the database somewhat myself (though using David’s linearts, that would be the first lineart style I would train it for) and the hardware were of course not that good.
It would have to be trained for every style (of line) it would work with, but I hoped to work around it with just retraining the last few layers instead of the whole network. So at first there would be just David’s linework, and then maybe I could get some other lineworks for a database for different styles. Maybe even allow users to train it themselves if it’s not too long (though it would require additional Python packages).
I haven’t finished it, first because I got into a difficult situation where the back propagation didn’t work correctly, and then I got a new shiny PC and I couldn’t make the plugin work there, so I got a bit discouraged. I hope I could continue working on it someday.
2 Likes
I just found a free AI, called Dream by Wombo. I kinda like it, the images are not perfect, but you can pick a style and add your own photos. Here are some pictures I generated:
3 Likes
Wow. I haven’t heard of this one. This is actually pretty cool and you can do it all on the website. This gives even more abstract results, but it kind of accomplishes the same idea of of providing rough directions you could go in.
3 Likes
I like it, I’ve also tried midjourney and it was definitely fun if nothing else. I haven’t been very successful with it so far but even the not sogreat images I got were still interesting ;).
I could definitely see a tool like this being used, I make small games and not that great with art so having something help me get some base concepts in 2d which I can then translate into 3d that would really great.
I’ve also beein checking photobashing and texturing and thought these areas could probably uses this kind of ai work quite a bit.
1 Like
AI’s definitely come a long way, it’s impressive technology regardless.
A couple handpicked renditions of Kiki from OpenAI’s Dalle-2
2 Likes
That was exactly what i thought when i saw that image, but at the same time it gives me a shot of inspiration. makes me want to make some arrangements and even tell a story.
I don’t know, in the end art is totally subjective
1 Like
Its a pretty mixed thing for me, it heavily depends on who is using it, how they are using it and what are they expecting to get back for it.
If someone uses one of these tools and then post it on their media, it is important to be transparant about the use of one of these programs.
If the person says “I used this AI program and it made this” then it is fine because they are being honest that they used a AI program for it. The end result can also bring up ideas when someone has artist block.
If the person says “I drew this in this digital art program all by had and its my original art” then that is a problem. Its similar to cases when a small artist gets their own art traced or just downright stolen by a bigger user who claims it is their own when it is not.
Using a result the AI made then editing it is also fine but like I said, be honest about it. Its not only just decent behavior but also common sense.
Thank You for coming to my TedTalk
3 Likes