AI Image Generators
Considering how some rich people have private collections of various original works of art, we see no reason why us ordinary folk could not have our own virtual collections or uniquely original AI generated images, — just a thought, but we are already accumulating our collection, so....
At first we were going to put some AI images, that we generated on various online generators, on a separate blog. However, after a quick Google search the AI indicated that X.com is a popular platform for sharing those images. So, we figured why fool with another blog, easier just to upload to X.
However decided to make a related post here on our main blog. With a few thoughts and examples.
Listed in alphabetical order.
__
A seemingly completely free image generator. Just to mention, since we did try it. It is somewhat slower than some of the others we tried, and the images (although actually quite good in composition), were somewhat blurry.
It does seem to allow an unlimited number of image creation without requiring an account. The images of free accounts are published publicly on their site, those of account holders apparently are not.
__
A good generator, with the many available options, but only allows generation of two hundred free images before you need to sign up for a paid account, which allows five hundred images per month. Their image generator can generate near photorealistic images and has options for size/dimensions and many different filters that can be applied to drastically change the styles of the image output.
This site does specifically state that you own the images that you create with their generator.
According to Google search, the AI text to image generator is free, but limited to the number allowed to generate (at any one time?). (?)
__
Did seem to be free, with no account, but now they allow twenty images before you need to create an account, which allows unlimited image creation.
According to Google search, the AI text to image generator is free for unlimited use. (?)
Although, not quite as photorealistic as some other generators, more like painted art, but some things like flowers do look pretty close.
(and it does seem to be getting better)
Some images seem a bit more fantasy world (and often a bit more colorful), such as castles, forests and island scenics, especially night scenes (and not quite as granular).
However, things like vehicles can seem a little outdated (think pre 1990s), but overall I think it generates some pretty nice images, reasonably fast.
Abstract art seems quite good compared to a few other generators we have tried (some were rather unremarkable). It seems to understand differing things like painting styles, media types and names of famous artists when used on the prompt (although sometimes, as with many AI generators, it may take two or three tries before it starts getting it right).
— and although not photorealistic, some of the images generated there do look nice on my four year old Android phone — and my older Windows 10 laptop
__
Nightcafé ?
Do not remember where we generated these, a few weeks ago (think the site was café something, or something café --- maybe Nightcafe?), but I think they turned out pretty nice. However, wherever it was, it stopped us at about twenty images before requiring us to create an account.
Whichever site it was, it seemed to be using tracking methods, other than just standard cookies. (?) 😒
__
Another recently discovered, which seems to be pretty good. However, only seems to allow a few free images to be generated before asking to create an account.
__
Seems totally free and unlimited in the numbers of images you can generate. However, there are quite a few pop-ups.
If you are using Firefox and blocker extensions you may need to play a bit with the settings. We had a bit of trouble until fiddling with one of the extensions (and disabling another). Perhaps you will have better luck, and it ís definitely worth a try.
The images are nearly photorealistic. Although it can be a bit slow (probably due to many users), and can seem a bit iffy at generating what you specify on the prompt (seems as if it gets easily confused, but does seem to be improving).
When we tried to regenerate a different image with a slightly different prompt, it generated the same image.
tip: Since the generator does not seem to automatically reseed. Try adding a short string of random letters and numbers to the end of your prompt, before attempting to regenerate with the same prompt.
Without reseeding, a duplicate image can be created using the same prompt. So, when posting, be aware that your image may not be unique, unless you use a long, elaborate description for the prompt. Or add a random string of characters to it. Even then the images can be quite similar. (unless you do a lot of editing)
It would be nice if the provider updated the generator to automatically reseed, but we are sure there are other reasons not to (increased user volume, higher energy consumption, …), which could negatively impact the usefulness. Or the providers ability to keep it free.
That said, this generator can create some excellent images.
__
Another recently discovered free and unlimited AI text to image generator we currently are playing with, which seems to be pretty good.
It is photorealistic, capable of generating very highly detailed images, with seemingly minimal hallucinations. Has several various aspect ratios and many different customizable image generator presets to choose from. This one is definitely worth checking out.
__
For something different, try adding words such as: muddy, rain, haze, smoke, spooky, floral, haunted, abandoned, desolate, sunny, lightning, moonlit and/or abstract, … to your prompts.
__
All the generators we have tried do seem susceptible to generating artifacts and mistakes. Some much worse than others. Some of the mistakes can actually be quite interesting.
Such as generating three-wheeler motorcycles with seemingly plausible, but technically ridiculous placements of the wheels.
Or full-length semi (tractor-trailer) with no rear axles on the truck, or extra axles on the trailer.
Other common mistakes seem to involve misshapen or too many arms, legs, fingers and hands. Some are worse for generating phantom text in the image, either gibberish or actual wording. One generator actually added the name of a serial killer (not hearsay, we actually seen that on one we generated).
--- and creating a jet engined dragster often seems quite difficult to get correct 😁
(update: try using "jet powered" 😃)
__
We are not so big on the whole “AI will improve the world” thing (nor destroy it), actually quite skeptical. However, not being artists, we do like the AI image generators. Although true, some real artist would hate them, and we can understand that. As well as the fact that they can be used maliciously to create fake images of real people.
But, from a computer hobbiest viewpoint, we think it is quite amazing (and amusing) how the AI can create the kinds of images that it does, so quickly, with such amazing detail.
Some might say to us, “Well, you musician types will be replaced by AI.” Our opinion; “Okay.” — our music is just a hobby
What fascinates me is how you can type in a few words at the prompt and the AI not only generates what you entered, but can be quite elaborate at the rest of the image, often with very meticulous detail, that you never specified.
Yeah, we know it learns from all sorts of imagery that it finds on the internet, but still ….
___
Some random hypothesis about some images that we created by using AI.
Cybernetic automobiles:
We like to generate automobiles with exposed mechanics, cables and hoses. Which we got the idea from steampunk style vehicles.
However, we decided to put a futuristic twist and imagined them as cybernetic. Admittedly, the exposed hardware would not be too practical in the real world, and a nightmare to clean. We imagined that there would be access panels, that would be installed for normal operation. Of course, with those installed the vehicle would look fairly normal. So, we decided that they are pictured with the panels removed, merely for display or maintenance.
As for cybernetic, whether they are machines with a biological brain, or an advanced CPU? You can decided, but we think the CPU is more likely.
And if all the extra cables and hoses were needed mostly for life support systems, it couldn't be very practical.
— but just think of the advanced weapons technology that could be tucked away within a vehicle sized machine
(just think about Terminator 3's T/X 😳)
___
— and now some painters are complaining how AI will flood the internet with inferior graphic art.
But hey, if they can create real works of art, better than the AI, — why complain?
--- and considering the vast amount of porn, music, and nearly everything else on the internet; there seems to be plenty of room for more 🤔
--- as usual, we will probably add more to this post, so check back occasionally
(we really should use thumbnails, but we are lazy)
Comments
Post a Comment